if_igb.c revision 254264
1/****************************************************************************** 2 3 Copyright (c) 2001-2013, Intel Corporation 4 All rights reserved. 5 6 Redistribution and use in source and binary forms, with or without 7 modification, are permitted provided that the following conditions are met: 8 9 1. Redistributions of source code must retain the above copyright notice, 10 this list of conditions and the following disclaimer. 11 12 2. Redistributions in binary form must reproduce the above copyright 13 notice, this list of conditions and the following disclaimer in the 14 documentation and/or other materials provided with the distribution. 15 16 3. Neither the name of the Intel Corporation nor the names of its 17 contributors may be used to endorse or promote products derived from 18 this software without specific prior written permission. 19 20 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 21 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 24 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 POSSIBILITY OF SUCH DAMAGE. 31 32******************************************************************************/ 33/*$FreeBSD: head/sys/dev/e1000/if_igb.c 254264 2013-08-13 00:25:39Z jfv $*/ 34 35 36#include "opt_inet.h" 37#include "opt_inet6.h" 38 39#ifdef HAVE_KERNEL_OPTION_HEADERS 40#include "opt_device_polling.h" 41#include "opt_altq.h" 42#endif 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#ifndef IGB_LEGACY_TX 47#include <sys/buf_ring.h> 48#endif 49#include <sys/bus.h> 50#include <sys/endian.h> 51#include <sys/kernel.h> 52#include <sys/kthread.h> 53#include <sys/malloc.h> 54#include <sys/mbuf.h> 55#include <sys/module.h> 56#include <sys/rman.h> 57#include <sys/socket.h> 58#include <sys/sockio.h> 59#include <sys/sysctl.h> 60#include <sys/taskqueue.h> 61#include <sys/eventhandler.h> 62#include <sys/pcpu.h> 63#include <sys/smp.h> 64#include <machine/smp.h> 65#include <machine/bus.h> 66#include <machine/resource.h> 67 68#include <net/bpf.h> 69#include <net/ethernet.h> 70#include <net/if.h> 71#include <net/if_arp.h> 72#include <net/if_dl.h> 73#include <net/if_media.h> 74 75#include <net/if_types.h> 76#include <net/if_vlan_var.h> 77 78#include <netinet/in_systm.h> 79#include <netinet/in.h> 80#include <netinet/if_ether.h> 81#include <netinet/ip.h> 82#include <netinet/ip6.h> 83#include <netinet/tcp.h> 84#include <netinet/tcp_lro.h> 85#include <netinet/udp.h> 86 87#include <machine/in_cksum.h> 88#include <dev/led/led.h> 89#include <dev/pci/pcivar.h> 90#include <dev/pci/pcireg.h> 91 92#include "e1000_api.h" 93#include "e1000_82575.h" 94#include "if_igb.h" 95 96/********************************************************************* 97 * Set this to one to display debug statistics 98 *********************************************************************/ 99int igb_display_debug_stats = 0; 100 101/********************************************************************* 102 * Driver version: 103 *********************************************************************/ 104char igb_driver_version[] = "version - 2.3.10"; 105 106 107/********************************************************************* 108 * PCI Device ID Table 109 * 110 * Used by probe to select devices to load on 111 * Last field stores an index into e1000_strings 112 * Last entry must be all 0s 113 * 114 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index } 115 *********************************************************************/ 116 117static igb_vendor_info_t igb_vendor_info_array[] = 118{ 119 { 0x8086, E1000_DEV_ID_82575EB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0}, 120 { 0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, 121 PCI_ANY_ID, PCI_ANY_ID, 0}, 122 { 0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, 123 PCI_ANY_ID, PCI_ANY_ID, 0}, 124 { 0x8086, E1000_DEV_ID_82576, PCI_ANY_ID, PCI_ANY_ID, 0}, 125 { 0x8086, E1000_DEV_ID_82576_NS, PCI_ANY_ID, PCI_ANY_ID, 0}, 126 { 0x8086, E1000_DEV_ID_82576_NS_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0}, 127 { 0x8086, E1000_DEV_ID_82576_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0}, 128 { 0x8086, E1000_DEV_ID_82576_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0}, 129 { 0x8086, E1000_DEV_ID_82576_SERDES_QUAD, 130 PCI_ANY_ID, PCI_ANY_ID, 0}, 131 { 0x8086, E1000_DEV_ID_82576_QUAD_COPPER, 132 PCI_ANY_ID, PCI_ANY_ID, 0}, 133 { 0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, 134 PCI_ANY_ID, PCI_ANY_ID, 0}, 135 { 0x8086, E1000_DEV_ID_82576_VF, PCI_ANY_ID, PCI_ANY_ID, 0}, 136 { 0x8086, E1000_DEV_ID_82580_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0}, 137 { 0x8086, E1000_DEV_ID_82580_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0}, 138 { 0x8086, E1000_DEV_ID_82580_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0}, 139 { 0x8086, E1000_DEV_ID_82580_SGMII, PCI_ANY_ID, PCI_ANY_ID, 0}, 140 { 0x8086, E1000_DEV_ID_82580_COPPER_DUAL, 141 PCI_ANY_ID, PCI_ANY_ID, 0}, 142 { 0x8086, E1000_DEV_ID_82580_QUAD_FIBER, 143 PCI_ANY_ID, PCI_ANY_ID, 0}, 144 { 0x8086, E1000_DEV_ID_DH89XXCC_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0}, 145 { 0x8086, E1000_DEV_ID_DH89XXCC_SGMII, PCI_ANY_ID, PCI_ANY_ID, 0}, 146 { 0x8086, E1000_DEV_ID_DH89XXCC_SFP, PCI_ANY_ID, PCI_ANY_ID, 0}, 147 { 0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, 148 PCI_ANY_ID, PCI_ANY_ID, 0}, 149 { 0x8086, E1000_DEV_ID_I350_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0}, 150 { 0x8086, E1000_DEV_ID_I350_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0}, 151 { 0x8086, E1000_DEV_ID_I350_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0}, 152 { 0x8086, E1000_DEV_ID_I350_SGMII, PCI_ANY_ID, PCI_ANY_ID, 0}, 153 { 0x8086, E1000_DEV_ID_I350_VF, PCI_ANY_ID, PCI_ANY_ID, 0}, 154 { 0x8086, E1000_DEV_ID_I210_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0}, 155 { 0x8086, E1000_DEV_ID_I210_COPPER_IT, PCI_ANY_ID, PCI_ANY_ID, 0}, 156 { 0x8086, E1000_DEV_ID_I210_COPPER_OEM1, 157 PCI_ANY_ID, PCI_ANY_ID, 0}, 158 { 0x8086, E1000_DEV_ID_I210_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0}, 159 { 0x8086, E1000_DEV_ID_I210_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0}, 160 { 0x8086, E1000_DEV_ID_I210_SGMII, PCI_ANY_ID, PCI_ANY_ID, 0}, 161 { 0x8086, E1000_DEV_ID_I211_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0}, 162 /* required last entry */ 163 { 0, 0, 0, 0, 0} 164}; 165 166/********************************************************************* 167 * Table of branding strings for all supported NICs. 168 *********************************************************************/ 169 170static char *igb_strings[] = { 171 "Intel(R) PRO/1000 Network Connection" 172}; 173 174/********************************************************************* 175 * Function prototypes 176 *********************************************************************/ 177static int igb_probe(device_t); 178static int igb_attach(device_t); 179static int igb_detach(device_t); 180static int igb_shutdown(device_t); 181static int igb_suspend(device_t); 182static int igb_resume(device_t); 183#ifndef IGB_LEGACY_TX 184static int igb_mq_start(struct ifnet *, struct mbuf *); 185static int igb_mq_start_locked(struct ifnet *, struct tx_ring *); 186static void igb_qflush(struct ifnet *); 187static void igb_deferred_mq_start(void *, int); 188#else 189static void igb_start(struct ifnet *); 190static void igb_start_locked(struct tx_ring *, struct ifnet *ifp); 191#endif 192static int igb_ioctl(struct ifnet *, u_long, caddr_t); 193static void igb_init(void *); 194static void igb_init_locked(struct adapter *); 195static void igb_stop(void *); 196static void igb_media_status(struct ifnet *, struct ifmediareq *); 197static int igb_media_change(struct ifnet *); 198static void igb_identify_hardware(struct adapter *); 199static int igb_allocate_pci_resources(struct adapter *); 200static int igb_allocate_msix(struct adapter *); 201static int igb_allocate_legacy(struct adapter *); 202static int igb_setup_msix(struct adapter *); 203static void igb_free_pci_resources(struct adapter *); 204static void igb_local_timer(void *); 205static void igb_reset(struct adapter *); 206static int igb_setup_interface(device_t, struct adapter *); 207static int igb_allocate_queues(struct adapter *); 208static void igb_configure_queues(struct adapter *); 209 210static int igb_allocate_transmit_buffers(struct tx_ring *); 211static void igb_setup_transmit_structures(struct adapter *); 212static void igb_setup_transmit_ring(struct tx_ring *); 213static void igb_initialize_transmit_units(struct adapter *); 214static void igb_free_transmit_structures(struct adapter *); 215static void igb_free_transmit_buffers(struct tx_ring *); 216 217static int igb_allocate_receive_buffers(struct rx_ring *); 218static int igb_setup_receive_structures(struct adapter *); 219static int igb_setup_receive_ring(struct rx_ring *); 220static void igb_initialize_receive_units(struct adapter *); 221static void igb_free_receive_structures(struct adapter *); 222static void igb_free_receive_buffers(struct rx_ring *); 223static void igb_free_receive_ring(struct rx_ring *); 224 225static void igb_enable_intr(struct adapter *); 226static void igb_disable_intr(struct adapter *); 227static void igb_update_stats_counters(struct adapter *); 228static bool igb_txeof(struct tx_ring *); 229 230static __inline void igb_rx_discard(struct rx_ring *, int); 231static __inline void igb_rx_input(struct rx_ring *, 232 struct ifnet *, struct mbuf *, u32); 233 234static bool igb_rxeof(struct igb_queue *, int, int *); 235static void igb_rx_checksum(u32, struct mbuf *, u32); 236static bool igb_tx_ctx_setup(struct tx_ring *, struct mbuf *); 237static bool igb_tso_setup(struct tx_ring *, struct mbuf *, int, 238 struct ip *, struct tcphdr *); 239static void igb_set_promisc(struct adapter *); 240static void igb_disable_promisc(struct adapter *); 241static void igb_set_multi(struct adapter *); 242static void igb_update_link_status(struct adapter *); 243static void igb_refresh_mbufs(struct rx_ring *, int); 244 245static void igb_register_vlan(void *, struct ifnet *, u16); 246static void igb_unregister_vlan(void *, struct ifnet *, u16); 247static void igb_setup_vlan_hw_support(struct adapter *); 248 249static int igb_xmit(struct tx_ring *, struct mbuf **); 250static int igb_dma_malloc(struct adapter *, bus_size_t, 251 struct igb_dma_alloc *, int); 252static void igb_dma_free(struct adapter *, struct igb_dma_alloc *); 253static int igb_sysctl_nvm_info(SYSCTL_HANDLER_ARGS); 254static void igb_print_nvm_info(struct adapter *); 255static int igb_is_valid_ether_addr(u8 *); 256static void igb_add_hw_stats(struct adapter *); 257 258static void igb_vf_init_stats(struct adapter *); 259static void igb_update_vf_stats_counters(struct adapter *); 260 261/* Management and WOL Support */ 262static void igb_init_manageability(struct adapter *); 263static void igb_release_manageability(struct adapter *); 264static void igb_get_hw_control(struct adapter *); 265static void igb_release_hw_control(struct adapter *); 266static void igb_enable_wakeup(device_t); 267static void igb_led_func(void *, int); 268 269static int igb_irq_fast(void *); 270static void igb_msix_que(void *); 271static void igb_msix_link(void *); 272static void igb_handle_que(void *context, int pending); 273static void igb_handle_link(void *context, int pending); 274static void igb_handle_link_locked(struct adapter *); 275 276static void igb_set_sysctl_value(struct adapter *, const char *, 277 const char *, int *, int); 278static int igb_set_flowcntl(SYSCTL_HANDLER_ARGS); 279static int igb_sysctl_dmac(SYSCTL_HANDLER_ARGS); 280static int igb_sysctl_eee(SYSCTL_HANDLER_ARGS); 281 282#ifdef DEVICE_POLLING 283static poll_handler_t igb_poll; 284#endif /* POLLING */ 285 286/********************************************************************* 287 * FreeBSD Device Interface Entry Points 288 *********************************************************************/ 289 290static device_method_t igb_methods[] = { 291 /* Device interface */ 292 DEVMETHOD(device_probe, igb_probe), 293 DEVMETHOD(device_attach, igb_attach), 294 DEVMETHOD(device_detach, igb_detach), 295 DEVMETHOD(device_shutdown, igb_shutdown), 296 DEVMETHOD(device_suspend, igb_suspend), 297 DEVMETHOD(device_resume, igb_resume), 298 DEVMETHOD_END 299}; 300 301static driver_t igb_driver = { 302 "igb", igb_methods, sizeof(struct adapter), 303}; 304 305static devclass_t igb_devclass; 306DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0); 307MODULE_DEPEND(igb, pci, 1, 1, 1); 308MODULE_DEPEND(igb, ether, 1, 1, 1); 309 310/********************************************************************* 311 * Tunable default values. 312 *********************************************************************/ 313 314static SYSCTL_NODE(_hw, OID_AUTO, igb, CTLFLAG_RD, 0, "IGB driver parameters"); 315 316/* Descriptor defaults */ 317static int igb_rxd = IGB_DEFAULT_RXD; 318static int igb_txd = IGB_DEFAULT_TXD; 319TUNABLE_INT("hw.igb.rxd", &igb_rxd); 320TUNABLE_INT("hw.igb.txd", &igb_txd); 321SYSCTL_INT(_hw_igb, OID_AUTO, rxd, CTLFLAG_RDTUN, &igb_rxd, 0, 322 "Number of receive descriptors per queue"); 323SYSCTL_INT(_hw_igb, OID_AUTO, txd, CTLFLAG_RDTUN, &igb_txd, 0, 324 "Number of transmit descriptors per queue"); 325 326/* 327** AIM: Adaptive Interrupt Moderation 328** which means that the interrupt rate 329** is varied over time based on the 330** traffic for that interrupt vector 331*/ 332static int igb_enable_aim = TRUE; 333TUNABLE_INT("hw.igb.enable_aim", &igb_enable_aim); 334SYSCTL_INT(_hw_igb, OID_AUTO, enable_aim, CTLFLAG_RW, &igb_enable_aim, 0, 335 "Enable adaptive interrupt moderation"); 336 337/* 338 * MSIX should be the default for best performance, 339 * but this allows it to be forced off for testing. 340 */ 341static int igb_enable_msix = 1; 342TUNABLE_INT("hw.igb.enable_msix", &igb_enable_msix); 343SYSCTL_INT(_hw_igb, OID_AUTO, enable_msix, CTLFLAG_RDTUN, &igb_enable_msix, 0, 344 "Enable MSI-X interrupts"); 345 346/* 347** Tuneable Interrupt rate 348*/ 349static int igb_max_interrupt_rate = 8000; 350TUNABLE_INT("hw.igb.max_interrupt_rate", &igb_max_interrupt_rate); 351SYSCTL_INT(_hw_igb, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN, 352 &igb_max_interrupt_rate, 0, "Maximum interrupts per second"); 353 354#if __FreeBSD_version >= 800000 355/* 356** Tuneable number of buffers in the buf-ring (drbr_xxx) 357*/ 358static int igb_buf_ring_size = IGB_BR_SIZE; 359TUNABLE_INT("hw.igb.buf_ring_size", &igb_buf_ring_size); 360SYSCTL_INT(_hw_igb, OID_AUTO, buf_ring_size, CTLFLAG_RDTUN, 361 &igb_buf_ring_size, 0, "Size of the bufring"); 362#endif 363 364/* 365** Header split causes the packet header to 366** be dma'd to a seperate mbuf from the payload. 367** this can have memory alignment benefits. But 368** another plus is that small packets often fit 369** into the header and thus use no cluster. Its 370** a very workload dependent type feature. 371*/ 372static int igb_header_split = FALSE; 373TUNABLE_INT("hw.igb.hdr_split", &igb_header_split); 374SYSCTL_INT(_hw_igb, OID_AUTO, header_split, CTLFLAG_RDTUN, &igb_header_split, 0, 375 "Enable receive mbuf header split"); 376 377/* 378** This will autoconfigure based on the 379** number of CPUs and max supported 380** MSIX messages if left at 0. 381*/ 382static int igb_num_queues = 0; 383TUNABLE_INT("hw.igb.num_queues", &igb_num_queues); 384SYSCTL_INT(_hw_igb, OID_AUTO, num_queues, CTLFLAG_RDTUN, &igb_num_queues, 0, 385 "Number of queues to configure, 0 indicates autoconfigure"); 386 387/* 388** Global variable to store last used CPU when binding queues 389** to CPUs in igb_allocate_msix. Starts at CPU_FIRST and increments when a 390** queue is bound to a cpu. 391*/ 392static int igb_last_bind_cpu = -1; 393 394/* How many packets rxeof tries to clean at a time */ 395static int igb_rx_process_limit = 100; 396TUNABLE_INT("hw.igb.rx_process_limit", &igb_rx_process_limit); 397SYSCTL_INT(_hw_igb, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN, 398 &igb_rx_process_limit, 0, 399 "Maximum number of received packets to process at a time, -1 means unlimited"); 400 401#ifdef DEV_NETMAP /* see ixgbe.c for details */ 402#include <dev/netmap/if_igb_netmap.h> 403#endif /* DEV_NETMAP */ 404/********************************************************************* 405 * Device identification routine 406 * 407 * igb_probe determines if the driver should be loaded on 408 * adapter based on PCI vendor/device id of the adapter. 409 * 410 * return BUS_PROBE_DEFAULT on success, positive on failure 411 *********************************************************************/ 412 413static int 414igb_probe(device_t dev) 415{ 416 char adapter_name[60]; 417 uint16_t pci_vendor_id = 0; 418 uint16_t pci_device_id = 0; 419 uint16_t pci_subvendor_id = 0; 420 uint16_t pci_subdevice_id = 0; 421 igb_vendor_info_t *ent; 422 423 INIT_DEBUGOUT("igb_probe: begin"); 424 425 pci_vendor_id = pci_get_vendor(dev); 426 if (pci_vendor_id != IGB_VENDOR_ID) 427 return (ENXIO); 428 429 pci_device_id = pci_get_device(dev); 430 pci_subvendor_id = pci_get_subvendor(dev); 431 pci_subdevice_id = pci_get_subdevice(dev); 432 433 ent = igb_vendor_info_array; 434 while (ent->vendor_id != 0) { 435 if ((pci_vendor_id == ent->vendor_id) && 436 (pci_device_id == ent->device_id) && 437 438 ((pci_subvendor_id == ent->subvendor_id) || 439 (ent->subvendor_id == PCI_ANY_ID)) && 440 441 ((pci_subdevice_id == ent->subdevice_id) || 442 (ent->subdevice_id == PCI_ANY_ID))) { 443 sprintf(adapter_name, "%s %s", 444 igb_strings[ent->index], 445 igb_driver_version); 446 device_set_desc_copy(dev, adapter_name); 447 return (BUS_PROBE_DEFAULT); 448 } 449 ent++; 450 } 451 452 return (ENXIO); 453} 454 455/********************************************************************* 456 * Device initialization routine 457 * 458 * The attach entry point is called when the driver is being loaded. 459 * This routine identifies the type of hardware, allocates all resources 460 * and initializes the hardware. 461 * 462 * return 0 on success, positive on failure 463 *********************************************************************/ 464 465static int 466igb_attach(device_t dev) 467{ 468 struct adapter *adapter; 469 int error = 0; 470 u16 eeprom_data; 471 472 INIT_DEBUGOUT("igb_attach: begin"); 473 474 if (resource_disabled("igb", device_get_unit(dev))) { 475 device_printf(dev, "Disabled by device hint\n"); 476 return (ENXIO); 477 } 478 479 adapter = device_get_softc(dev); 480 adapter->dev = adapter->osdep.dev = dev; 481 IGB_CORE_LOCK_INIT(adapter, device_get_nameunit(dev)); 482 483 /* SYSCTL stuff */ 484 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), 485 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), 486 OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0, 487 igb_sysctl_nvm_info, "I", "NVM Information"); 488 489 igb_set_sysctl_value(adapter, "enable_aim", 490 "Interrupt Moderation", &adapter->enable_aim, 491 igb_enable_aim); 492 493 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), 494 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), 495 OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, 496 adapter, 0, igb_set_flowcntl, "I", "Flow Control"); 497 498 callout_init_mtx(&adapter->timer, &adapter->core_mtx, 0); 499 500 /* Determine hardware and mac info */ 501 igb_identify_hardware(adapter); 502 503 /* Setup PCI resources */ 504 if (igb_allocate_pci_resources(adapter)) { 505 device_printf(dev, "Allocation of PCI resources failed\n"); 506 error = ENXIO; 507 goto err_pci; 508 } 509 510 /* Do Shared Code initialization */ 511 if (e1000_setup_init_funcs(&adapter->hw, TRUE)) { 512 device_printf(dev, "Setup of Shared code failed\n"); 513 error = ENXIO; 514 goto err_pci; 515 } 516 517 e1000_get_bus_info(&adapter->hw); 518 519 /* Sysctl for limiting the amount of work done in the taskqueue */ 520 igb_set_sysctl_value(adapter, "rx_processing_limit", 521 "max number of rx packets to process", 522 &adapter->rx_process_limit, igb_rx_process_limit); 523 524 /* 525 * Validate number of transmit and receive descriptors. It 526 * must not exceed hardware maximum, and must be multiple 527 * of E1000_DBA_ALIGN. 528 */ 529 if (((igb_txd * sizeof(struct e1000_tx_desc)) % IGB_DBA_ALIGN) != 0 || 530 (igb_txd > IGB_MAX_TXD) || (igb_txd < IGB_MIN_TXD)) { 531 device_printf(dev, "Using %d TX descriptors instead of %d!\n", 532 IGB_DEFAULT_TXD, igb_txd); 533 adapter->num_tx_desc = IGB_DEFAULT_TXD; 534 } else 535 adapter->num_tx_desc = igb_txd; 536 if (((igb_rxd * sizeof(struct e1000_rx_desc)) % IGB_DBA_ALIGN) != 0 || 537 (igb_rxd > IGB_MAX_RXD) || (igb_rxd < IGB_MIN_RXD)) { 538 device_printf(dev, "Using %d RX descriptors instead of %d!\n", 539 IGB_DEFAULT_RXD, igb_rxd); 540 adapter->num_rx_desc = IGB_DEFAULT_RXD; 541 } else 542 adapter->num_rx_desc = igb_rxd; 543 544 adapter->hw.mac.autoneg = DO_AUTO_NEG; 545 adapter->hw.phy.autoneg_wait_to_complete = FALSE; 546 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT; 547 548 /* Copper options */ 549 if (adapter->hw.phy.media_type == e1000_media_type_copper) { 550 adapter->hw.phy.mdix = AUTO_ALL_MODES; 551 adapter->hw.phy.disable_polarity_correction = FALSE; 552 adapter->hw.phy.ms_type = IGB_MASTER_SLAVE; 553 } 554 555 /* 556 * Set the frame limits assuming 557 * standard ethernet sized frames. 558 */ 559 adapter->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE; 560 adapter->min_frame_size = ETH_ZLEN + ETHERNET_FCS_SIZE; 561 562 /* 563 ** Allocate and Setup Queues 564 */ 565 if (igb_allocate_queues(adapter)) { 566 error = ENOMEM; 567 goto err_pci; 568 } 569 570 /* Allocate the appropriate stats memory */ 571 if (adapter->vf_ifp) { 572 adapter->stats = 573 (struct e1000_vf_stats *)malloc(sizeof \ 574 (struct e1000_vf_stats), M_DEVBUF, M_NOWAIT | M_ZERO); 575 igb_vf_init_stats(adapter); 576 } else 577 adapter->stats = 578 (struct e1000_hw_stats *)malloc(sizeof \ 579 (struct e1000_hw_stats), M_DEVBUF, M_NOWAIT | M_ZERO); 580 if (adapter->stats == NULL) { 581 device_printf(dev, "Can not allocate stats memory\n"); 582 error = ENOMEM; 583 goto err_late; 584 } 585 586 /* Allocate multicast array memory. */ 587 adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN * 588 MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT); 589 if (adapter->mta == NULL) { 590 device_printf(dev, "Can not allocate multicast setup array\n"); 591 error = ENOMEM; 592 goto err_late; 593 } 594 595 /* Some adapter-specific advanced features */ 596 if (adapter->hw.mac.type >= e1000_i350) { 597 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), 598 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), 599 OID_AUTO, "dmac", CTLTYPE_INT|CTLFLAG_RW, 600 adapter, 0, igb_sysctl_dmac, "I", "DMA Coalesce"); 601 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), 602 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), 603 OID_AUTO, "eee_disabled", CTLTYPE_INT|CTLFLAG_RW, 604 adapter, 0, igb_sysctl_eee, "I", 605 "Disable Energy Efficient Ethernet"); 606 if (adapter->hw.phy.media_type == e1000_media_type_copper) 607 e1000_set_eee_i350(&adapter->hw); 608 } 609 610 /* 611 ** Start from a known state, this is 612 ** important in reading the nvm and 613 ** mac from that. 614 */ 615 e1000_reset_hw(&adapter->hw); 616 617 /* Make sure we have a good EEPROM before we read from it */ 618 if (((adapter->hw.mac.type != e1000_i210) && 619 (adapter->hw.mac.type != e1000_i211)) && 620 (e1000_validate_nvm_checksum(&adapter->hw) < 0)) { 621 /* 622 ** Some PCI-E parts fail the first check due to 623 ** the link being in sleep state, call it again, 624 ** if it fails a second time its a real issue. 625 */ 626 if (e1000_validate_nvm_checksum(&adapter->hw) < 0) { 627 device_printf(dev, 628 "The EEPROM Checksum Is Not Valid\n"); 629 error = EIO; 630 goto err_late; 631 } 632 } 633 634 /* 635 ** Copy the permanent MAC address out of the EEPROM 636 */ 637 if (e1000_read_mac_addr(&adapter->hw) < 0) { 638 device_printf(dev, "EEPROM read error while reading MAC" 639 " address\n"); 640 error = EIO; 641 goto err_late; 642 } 643 /* Check its sanity */ 644 if (!igb_is_valid_ether_addr(adapter->hw.mac.addr)) { 645 device_printf(dev, "Invalid MAC address\n"); 646 error = EIO; 647 goto err_late; 648 } 649 650 /* Setup OS specific network interface */ 651 if (igb_setup_interface(dev, adapter) != 0) 652 goto err_late; 653 654 /* Now get a good starting state */ 655 igb_reset(adapter); 656 657 /* Initialize statistics */ 658 igb_update_stats_counters(adapter); 659 660 adapter->hw.mac.get_link_status = 1; 661 igb_update_link_status(adapter); 662 663 /* Indicate SOL/IDER usage */ 664 if (e1000_check_reset_block(&adapter->hw)) 665 device_printf(dev, 666 "PHY reset is blocked due to SOL/IDER session.\n"); 667 668 /* Determine if we have to control management hardware */ 669 adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw); 670 671 /* 672 * Setup Wake-on-Lan 673 */ 674 /* APME bit in EEPROM is mapped to WUC.APME */ 675 eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC) & E1000_WUC_APME; 676 if (eeprom_data) 677 adapter->wol = E1000_WUFC_MAG; 678 679 /* Register for VLAN events */ 680 adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config, 681 igb_register_vlan, adapter, EVENTHANDLER_PRI_FIRST); 682 adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig, 683 igb_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST); 684 685 igb_add_hw_stats(adapter); 686 687 /* Tell the stack that the interface is not active */ 688 adapter->ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 689 adapter->ifp->if_drv_flags |= IFF_DRV_OACTIVE; 690 691 adapter->led_dev = led_create(igb_led_func, adapter, 692 device_get_nameunit(dev)); 693 694 /* 695 ** Configure Interrupts 696 */ 697 if ((adapter->msix > 1) && (igb_enable_msix)) 698 error = igb_allocate_msix(adapter); 699 else /* MSI or Legacy */ 700 error = igb_allocate_legacy(adapter); 701 if (error) 702 goto err_late; 703 704#ifdef DEV_NETMAP 705 igb_netmap_attach(adapter); 706#endif /* DEV_NETMAP */ 707 INIT_DEBUGOUT("igb_attach: end"); 708 709 return (0); 710 711err_late: 712 igb_detach(dev); 713 igb_free_transmit_structures(adapter); 714 igb_free_receive_structures(adapter); 715 igb_release_hw_control(adapter); 716err_pci: 717 igb_free_pci_resources(adapter); 718 if (adapter->ifp != NULL) 719 if_free(adapter->ifp); 720 free(adapter->mta, M_DEVBUF); 721 IGB_CORE_LOCK_DESTROY(adapter); 722 723 return (error); 724} 725 726/********************************************************************* 727 * Device removal routine 728 * 729 * The detach entry point is called when the driver is being removed. 730 * This routine stops the adapter and deallocates all the resources 731 * that were allocated for driver operation. 732 * 733 * return 0 on success, positive on failure 734 *********************************************************************/ 735 736static int 737igb_detach(device_t dev) 738{ 739 struct adapter *adapter = device_get_softc(dev); 740 struct ifnet *ifp = adapter->ifp; 741 742 INIT_DEBUGOUT("igb_detach: begin"); 743 744 /* Make sure VLANS are not using driver */ 745 if (adapter->ifp->if_vlantrunk != NULL) { 746 device_printf(dev,"Vlan in use, detach first\n"); 747 return (EBUSY); 748 } 749 750 ether_ifdetach(adapter->ifp); 751 752 if (adapter->led_dev != NULL) 753 led_destroy(adapter->led_dev); 754 755#ifdef DEVICE_POLLING 756 if (ifp->if_capenable & IFCAP_POLLING) 757 ether_poll_deregister(ifp); 758#endif 759 760 IGB_CORE_LOCK(adapter); 761 adapter->in_detach = 1; 762 igb_stop(adapter); 763 IGB_CORE_UNLOCK(adapter); 764 765 e1000_phy_hw_reset(&adapter->hw); 766 767 /* Give control back to firmware */ 768 igb_release_manageability(adapter); 769 igb_release_hw_control(adapter); 770 771 if (adapter->wol) { 772 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN); 773 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol); 774 igb_enable_wakeup(dev); 775 } 776 777 /* Unregister VLAN events */ 778 if (adapter->vlan_attach != NULL) 779 EVENTHANDLER_DEREGISTER(vlan_config, adapter->vlan_attach); 780 if (adapter->vlan_detach != NULL) 781 EVENTHANDLER_DEREGISTER(vlan_unconfig, adapter->vlan_detach); 782 783 callout_drain(&adapter->timer); 784 785#ifdef DEV_NETMAP 786 netmap_detach(adapter->ifp); 787#endif /* DEV_NETMAP */ 788 igb_free_pci_resources(adapter); 789 bus_generic_detach(dev); 790 if_free(ifp); 791 792 igb_free_transmit_structures(adapter); 793 igb_free_receive_structures(adapter); 794 if (adapter->mta != NULL) 795 free(adapter->mta, M_DEVBUF); 796 797 IGB_CORE_LOCK_DESTROY(adapter); 798 799 return (0); 800} 801 802/********************************************************************* 803 * 804 * Shutdown entry point 805 * 806 **********************************************************************/ 807 808static int 809igb_shutdown(device_t dev) 810{ 811 return igb_suspend(dev); 812} 813 814/* 815 * Suspend/resume device methods. 816 */ 817static int 818igb_suspend(device_t dev) 819{ 820 struct adapter *adapter = device_get_softc(dev); 821 822 IGB_CORE_LOCK(adapter); 823 824 igb_stop(adapter); 825 826 igb_release_manageability(adapter); 827 igb_release_hw_control(adapter); 828 829 if (adapter->wol) { 830 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN); 831 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol); 832 igb_enable_wakeup(dev); 833 } 834 835 IGB_CORE_UNLOCK(adapter); 836 837 return bus_generic_suspend(dev); 838} 839 840static int 841igb_resume(device_t dev) 842{ 843 struct adapter *adapter = device_get_softc(dev); 844 struct tx_ring *txr = adapter->tx_rings; 845 struct ifnet *ifp = adapter->ifp; 846 847 IGB_CORE_LOCK(adapter); 848 igb_init_locked(adapter); 849 igb_init_manageability(adapter); 850 851 if ((ifp->if_flags & IFF_UP) && 852 (ifp->if_drv_flags & IFF_DRV_RUNNING) && adapter->link_active) { 853 for (int i = 0; i < adapter->num_queues; i++, txr++) { 854 IGB_TX_LOCK(txr); 855#ifndef IGB_LEGACY_TX 856 /* Process the stack queue only if not depleted */ 857 if (((txr->queue_status & IGB_QUEUE_DEPLETED) == 0) && 858 !drbr_empty(ifp, txr->br)) 859 igb_mq_start_locked(ifp, txr); 860#else 861 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 862 igb_start_locked(txr, ifp); 863#endif 864 IGB_TX_UNLOCK(txr); 865 } 866 } 867 IGB_CORE_UNLOCK(adapter); 868 869 return bus_generic_resume(dev); 870} 871 872 873#ifdef IGB_LEGACY_TX 874 875/********************************************************************* 876 * Transmit entry point 877 * 878 * igb_start is called by the stack to initiate a transmit. 879 * The driver will remain in this routine as long as there are 880 * packets to transmit and transmit resources are available. 881 * In case resources are not available stack is notified and 882 * the packet is requeued. 883 **********************************************************************/ 884 885static void 886igb_start_locked(struct tx_ring *txr, struct ifnet *ifp) 887{ 888 struct adapter *adapter = ifp->if_softc; 889 struct mbuf *m_head; 890 891 IGB_TX_LOCK_ASSERT(txr); 892 893 if ((ifp->if_drv_flags & (IFF_DRV_RUNNING|IFF_DRV_OACTIVE)) != 894 IFF_DRV_RUNNING) 895 return; 896 if (!adapter->link_active) 897 return; 898 899 /* Call cleanup if number of TX descriptors low */ 900 if (txr->tx_avail <= IGB_TX_CLEANUP_THRESHOLD) 901 igb_txeof(txr); 902 903 while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) { 904 if (txr->tx_avail <= IGB_MAX_SCATTER) { 905 txr->queue_status |= IGB_QUEUE_DEPLETED; 906 break; 907 } 908 IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head); 909 if (m_head == NULL) 910 break; 911 /* 912 * Encapsulation can modify our pointer, and or make it 913 * NULL on failure. In that event, we can't requeue. 914 */ 915 if (igb_xmit(txr, &m_head)) { 916 if (m_head != NULL) 917 IFQ_DRV_PREPEND(&ifp->if_snd, m_head); 918 if (txr->tx_avail <= IGB_MAX_SCATTER) 919 txr->queue_status |= IGB_QUEUE_DEPLETED; 920 break; 921 } 922 923 /* Send a copy of the frame to the BPF listener */ 924 ETHER_BPF_MTAP(ifp, m_head); 925 926 /* Set watchdog on */ 927 txr->watchdog_time = ticks; 928 txr->queue_status |= IGB_QUEUE_WORKING; 929 } 930} 931 932/* 933 * Legacy TX driver routine, called from the 934 * stack, always uses tx[0], and spins for it. 935 * Should not be used with multiqueue tx 936 */ 937static void 938igb_start(struct ifnet *ifp) 939{ 940 struct adapter *adapter = ifp->if_softc; 941 struct tx_ring *txr = adapter->tx_rings; 942 943 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 944 IGB_TX_LOCK(txr); 945 igb_start_locked(txr, ifp); 946 IGB_TX_UNLOCK(txr); 947 } 948 return; 949} 950 951#else /* ~IGB_LEGACY_TX */ 952 953/* 954** Multiqueue Transmit Entry: 955** quick turnaround to the stack 956** 957*/ 958static int 959igb_mq_start(struct ifnet *ifp, struct mbuf *m) 960{ 961 struct adapter *adapter = ifp->if_softc; 962 struct igb_queue *que; 963 struct tx_ring *txr; 964 int i, err = 0; 965 966 /* Which queue to use */ 967 if ((m->m_flags & M_FLOWID) != 0) 968 i = m->m_pkthdr.flowid % adapter->num_queues; 969 else 970 i = curcpu % adapter->num_queues; 971 txr = &adapter->tx_rings[i]; 972 que = &adapter->queues[i]; 973 974 err = drbr_enqueue(ifp, txr->br, m); 975 if (err) 976 return (err); 977 if (IGB_TX_TRYLOCK(txr)) { 978 err = igb_mq_start_locked(ifp, txr); 979 IGB_TX_UNLOCK(txr); 980 } else 981 taskqueue_enqueue(que->tq, &txr->txq_task); 982 983 return (err); 984} 985 986static int 987igb_mq_start_locked(struct ifnet *ifp, struct tx_ring *txr) 988{ 989 struct adapter *adapter = txr->adapter; 990 struct mbuf *next; 991 int err = 0, enq; 992 993 IGB_TX_LOCK_ASSERT(txr); 994 995 if (((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) || 996 adapter->link_active == 0) 997 return (ENETDOWN); 998 999 enq = 0; 1000 1001 /* Process the queue */ 1002 while ((next = drbr_peek(ifp, txr->br)) != NULL) { 1003 if ((err = igb_xmit(txr, &next)) != 0) { 1004 if (next == NULL) { 1005 /* It was freed, move forward */ 1006 drbr_advance(ifp, txr->br); 1007 } else { 1008 /* 1009 * Still have one left, it may not be 1010 * the same since the transmit function 1011 * may have changed it. 1012 */ 1013 drbr_putback(ifp, txr->br, next); 1014 } 1015 break; 1016 } 1017 drbr_advance(ifp, txr->br); 1018 enq++; 1019 ifp->if_obytes += next->m_pkthdr.len; 1020 if (next->m_flags & M_MCAST) 1021 ifp->if_omcasts++; 1022 ETHER_BPF_MTAP(ifp, next); 1023 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 1024 break; 1025 } 1026 if (enq > 0) { 1027 /* Set the watchdog */ 1028 txr->queue_status |= IGB_QUEUE_WORKING; 1029 txr->watchdog_time = ticks; 1030 } 1031 if (txr->tx_avail <= IGB_TX_CLEANUP_THRESHOLD) 1032 igb_txeof(txr); 1033 if (txr->tx_avail <= IGB_MAX_SCATTER) 1034 txr->queue_status |= IGB_QUEUE_DEPLETED; 1035 return (err); 1036} 1037 1038/* 1039 * Called from a taskqueue to drain queued transmit packets. 1040 */ 1041static void 1042igb_deferred_mq_start(void *arg, int pending) 1043{ 1044 struct tx_ring *txr = arg; 1045 struct adapter *adapter = txr->adapter; 1046 struct ifnet *ifp = adapter->ifp; 1047 1048 IGB_TX_LOCK(txr); 1049 if (!drbr_empty(ifp, txr->br)) 1050 igb_mq_start_locked(ifp, txr); 1051 IGB_TX_UNLOCK(txr); 1052} 1053 1054/* 1055** Flush all ring buffers 1056*/ 1057static void 1058igb_qflush(struct ifnet *ifp) 1059{ 1060 struct adapter *adapter = ifp->if_softc; 1061 struct tx_ring *txr = adapter->tx_rings; 1062 struct mbuf *m; 1063 1064 for (int i = 0; i < adapter->num_queues; i++, txr++) { 1065 IGB_TX_LOCK(txr); 1066 while ((m = buf_ring_dequeue_sc(txr->br)) != NULL) 1067 m_freem(m); 1068 IGB_TX_UNLOCK(txr); 1069 } 1070 if_qflush(ifp); 1071} 1072#endif /* ~IGB_LEGACY_TX */ 1073 1074/********************************************************************* 1075 * Ioctl entry point 1076 * 1077 * igb_ioctl is called when the user wants to configure the 1078 * interface. 1079 * 1080 * return 0 on success, positive on failure 1081 **********************************************************************/ 1082 1083static int 1084igb_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1085{ 1086 struct adapter *adapter = ifp->if_softc; 1087 struct ifreq *ifr = (struct ifreq *)data; 1088#if defined(INET) || defined(INET6) 1089 struct ifaddr *ifa = (struct ifaddr *)data; 1090#endif 1091 bool avoid_reset = FALSE; 1092 int error = 0; 1093 1094 if (adapter->in_detach) 1095 return (error); 1096 1097 switch (command) { 1098 case SIOCSIFADDR: 1099#ifdef INET 1100 if (ifa->ifa_addr->sa_family == AF_INET) 1101 avoid_reset = TRUE; 1102#endif 1103#ifdef INET6 1104 if (ifa->ifa_addr->sa_family == AF_INET6) 1105 avoid_reset = TRUE; 1106#endif 1107 /* 1108 ** Calling init results in link renegotiation, 1109 ** so we avoid doing it when possible. 1110 */ 1111 if (avoid_reset) { 1112 ifp->if_flags |= IFF_UP; 1113 if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) 1114 igb_init(adapter); 1115#ifdef INET 1116 if (!(ifp->if_flags & IFF_NOARP)) 1117 arp_ifinit(ifp, ifa); 1118#endif 1119 } else 1120 error = ether_ioctl(ifp, command, data); 1121 break; 1122 case SIOCSIFMTU: 1123 { 1124 int max_frame_size; 1125 1126 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)"); 1127 1128 IGB_CORE_LOCK(adapter); 1129 max_frame_size = 9234; 1130 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN - 1131 ETHER_CRC_LEN) { 1132 IGB_CORE_UNLOCK(adapter); 1133 error = EINVAL; 1134 break; 1135 } 1136 1137 ifp->if_mtu = ifr->ifr_mtu; 1138 adapter->max_frame_size = 1139 ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN; 1140 igb_init_locked(adapter); 1141 IGB_CORE_UNLOCK(adapter); 1142 break; 1143 } 1144 case SIOCSIFFLAGS: 1145 IOCTL_DEBUGOUT("ioctl rcv'd:\ 1146 SIOCSIFFLAGS (Set Interface Flags)"); 1147 IGB_CORE_LOCK(adapter); 1148 if (ifp->if_flags & IFF_UP) { 1149 if ((ifp->if_drv_flags & IFF_DRV_RUNNING)) { 1150 if ((ifp->if_flags ^ adapter->if_flags) & 1151 (IFF_PROMISC | IFF_ALLMULTI)) { 1152 igb_disable_promisc(adapter); 1153 igb_set_promisc(adapter); 1154 } 1155 } else 1156 igb_init_locked(adapter); 1157 } else 1158 if (ifp->if_drv_flags & IFF_DRV_RUNNING) 1159 igb_stop(adapter); 1160 adapter->if_flags = ifp->if_flags; 1161 IGB_CORE_UNLOCK(adapter); 1162 break; 1163 case SIOCADDMULTI: 1164 case SIOCDELMULTI: 1165 IOCTL_DEBUGOUT("ioctl rcv'd: SIOC(ADD|DEL)MULTI"); 1166 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 1167 IGB_CORE_LOCK(adapter); 1168 igb_disable_intr(adapter); 1169 igb_set_multi(adapter); 1170#ifdef DEVICE_POLLING 1171 if (!(ifp->if_capenable & IFCAP_POLLING)) 1172#endif 1173 igb_enable_intr(adapter); 1174 IGB_CORE_UNLOCK(adapter); 1175 } 1176 break; 1177 case SIOCSIFMEDIA: 1178 /* Check SOL/IDER usage */ 1179 IGB_CORE_LOCK(adapter); 1180 if (e1000_check_reset_block(&adapter->hw)) { 1181 IGB_CORE_UNLOCK(adapter); 1182 device_printf(adapter->dev, "Media change is" 1183 " blocked due to SOL/IDER session.\n"); 1184 break; 1185 } 1186 IGB_CORE_UNLOCK(adapter); 1187 case SIOCGIFMEDIA: 1188 IOCTL_DEBUGOUT("ioctl rcv'd: \ 1189 SIOCxIFMEDIA (Get/Set Interface Media)"); 1190 error = ifmedia_ioctl(ifp, ifr, &adapter->media, command); 1191 break; 1192 case SIOCSIFCAP: 1193 { 1194 int mask, reinit; 1195 1196 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFCAP (Set Capabilities)"); 1197 reinit = 0; 1198 mask = ifr->ifr_reqcap ^ ifp->if_capenable; 1199#ifdef DEVICE_POLLING 1200 if (mask & IFCAP_POLLING) { 1201 if (ifr->ifr_reqcap & IFCAP_POLLING) { 1202 error = ether_poll_register(igb_poll, ifp); 1203 if (error) 1204 return (error); 1205 IGB_CORE_LOCK(adapter); 1206 igb_disable_intr(adapter); 1207 ifp->if_capenable |= IFCAP_POLLING; 1208 IGB_CORE_UNLOCK(adapter); 1209 } else { 1210 error = ether_poll_deregister(ifp); 1211 /* Enable interrupt even in error case */ 1212 IGB_CORE_LOCK(adapter); 1213 igb_enable_intr(adapter); 1214 ifp->if_capenable &= ~IFCAP_POLLING; 1215 IGB_CORE_UNLOCK(adapter); 1216 } 1217 } 1218#endif 1219 if (mask & IFCAP_HWCSUM) { 1220 ifp->if_capenable ^= IFCAP_HWCSUM; 1221 reinit = 1; 1222 } 1223 if (mask & IFCAP_TSO4) { 1224 ifp->if_capenable ^= IFCAP_TSO4; 1225 reinit = 1; 1226 } 1227 if (mask & IFCAP_VLAN_HWTAGGING) { 1228 ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; 1229 reinit = 1; 1230 } 1231 if (mask & IFCAP_VLAN_HWFILTER) { 1232 ifp->if_capenable ^= IFCAP_VLAN_HWFILTER; 1233 reinit = 1; 1234 } 1235 if (mask & IFCAP_VLAN_HWTSO) { 1236 ifp->if_capenable ^= IFCAP_VLAN_HWTSO; 1237 reinit = 1; 1238 } 1239 if (mask & IFCAP_LRO) { 1240 ifp->if_capenable ^= IFCAP_LRO; 1241 reinit = 1; 1242 } 1243 if (reinit && (ifp->if_drv_flags & IFF_DRV_RUNNING)) 1244 igb_init(adapter); 1245 VLAN_CAPABILITIES(ifp); 1246 break; 1247 } 1248 1249 default: 1250 error = ether_ioctl(ifp, command, data); 1251 break; 1252 } 1253 1254 return (error); 1255} 1256 1257 1258/********************************************************************* 1259 * Init entry point 1260 * 1261 * This routine is used in two ways. It is used by the stack as 1262 * init entry point in network interface structure. It is also used 1263 * by the driver as a hw/sw initialization routine to get to a 1264 * consistent state. 1265 * 1266 * return 0 on success, positive on failure 1267 **********************************************************************/ 1268 1269static void 1270igb_init_locked(struct adapter *adapter) 1271{ 1272 struct ifnet *ifp = adapter->ifp; 1273 device_t dev = adapter->dev; 1274 1275 INIT_DEBUGOUT("igb_init: begin"); 1276 1277 IGB_CORE_LOCK_ASSERT(adapter); 1278 1279 igb_disable_intr(adapter); 1280 callout_stop(&adapter->timer); 1281 1282 /* Get the latest mac address, User can use a LAA */ 1283 bcopy(IF_LLADDR(adapter->ifp), adapter->hw.mac.addr, 1284 ETHER_ADDR_LEN); 1285 1286 /* Put the address into the Receive Address Array */ 1287 e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0); 1288 1289 igb_reset(adapter); 1290 igb_update_link_status(adapter); 1291 1292 E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN); 1293 1294 /* Set hardware offload abilities */ 1295 ifp->if_hwassist = 0; 1296 if (ifp->if_capenable & IFCAP_TXCSUM) { 1297 ifp->if_hwassist |= (CSUM_TCP | CSUM_UDP); 1298#if __FreeBSD_version >= 800000 1299 if (adapter->hw.mac.type == e1000_82576) 1300 ifp->if_hwassist |= CSUM_SCTP; 1301#endif 1302 } 1303 1304 if (ifp->if_capenable & IFCAP_TSO4) 1305 ifp->if_hwassist |= CSUM_TSO; 1306 1307 /* Configure for OS presence */ 1308 igb_init_manageability(adapter); 1309 1310 /* Prepare transmit descriptors and buffers */ 1311 igb_setup_transmit_structures(adapter); 1312 igb_initialize_transmit_units(adapter); 1313 1314 /* Setup Multicast table */ 1315 igb_set_multi(adapter); 1316 1317 /* 1318 ** Figure out the desired mbuf pool 1319 ** for doing jumbo/packetsplit 1320 */ 1321 if (adapter->max_frame_size <= 2048) 1322 adapter->rx_mbuf_sz = MCLBYTES; 1323 else if (adapter->max_frame_size <= 4096) 1324 adapter->rx_mbuf_sz = MJUMPAGESIZE; 1325 else 1326 adapter->rx_mbuf_sz = MJUM9BYTES; 1327 1328 /* Prepare receive descriptors and buffers */ 1329 if (igb_setup_receive_structures(adapter)) { 1330 device_printf(dev, "Could not setup receive structures\n"); 1331 return; 1332 } 1333 igb_initialize_receive_units(adapter); 1334 1335 /* Enable VLAN support */ 1336 if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) 1337 igb_setup_vlan_hw_support(adapter); 1338 1339 /* Don't lose promiscuous settings */ 1340 igb_set_promisc(adapter); 1341 1342 ifp->if_drv_flags |= IFF_DRV_RUNNING; 1343 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 1344 1345 callout_reset(&adapter->timer, hz, igb_local_timer, adapter); 1346 e1000_clear_hw_cntrs_base_generic(&adapter->hw); 1347 1348 if (adapter->msix > 1) /* Set up queue routing */ 1349 igb_configure_queues(adapter); 1350 1351 /* this clears any pending interrupts */ 1352 E1000_READ_REG(&adapter->hw, E1000_ICR); 1353#ifdef DEVICE_POLLING 1354 /* 1355 * Only enable interrupts if we are not polling, make sure 1356 * they are off otherwise. 1357 */ 1358 if (ifp->if_capenable & IFCAP_POLLING) 1359 igb_disable_intr(adapter); 1360 else 1361#endif /* DEVICE_POLLING */ 1362 { 1363 igb_enable_intr(adapter); 1364 E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC); 1365 } 1366 1367 /* Set Energy Efficient Ethernet */ 1368 if (adapter->hw.phy.media_type == e1000_media_type_copper) 1369 e1000_set_eee_i350(&adapter->hw); 1370} 1371 1372static void 1373igb_init(void *arg) 1374{ 1375 struct adapter *adapter = arg; 1376 1377 IGB_CORE_LOCK(adapter); 1378 igb_init_locked(adapter); 1379 IGB_CORE_UNLOCK(adapter); 1380} 1381 1382 1383static void 1384igb_handle_que(void *context, int pending) 1385{ 1386 struct igb_queue *que = context; 1387 struct adapter *adapter = que->adapter; 1388 struct tx_ring *txr = que->txr; 1389 struct ifnet *ifp = adapter->ifp; 1390 1391 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 1392 bool more; 1393 1394 more = igb_rxeof(que, adapter->rx_process_limit, NULL); 1395 1396 IGB_TX_LOCK(txr); 1397 igb_txeof(txr); 1398#ifndef IGB_LEGACY_TX 1399 /* Process the stack queue only if not depleted */ 1400 if (((txr->queue_status & IGB_QUEUE_DEPLETED) == 0) && 1401 !drbr_empty(ifp, txr->br)) 1402 igb_mq_start_locked(ifp, txr); 1403#else 1404 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 1405 igb_start_locked(txr, ifp); 1406#endif 1407 IGB_TX_UNLOCK(txr); 1408 /* Do we need another? */ 1409 if (more) { 1410 taskqueue_enqueue(que->tq, &que->que_task); 1411 return; 1412 } 1413 } 1414 1415#ifdef DEVICE_POLLING 1416 if (ifp->if_capenable & IFCAP_POLLING) 1417 return; 1418#endif 1419 /* Reenable this interrupt */ 1420 if (que->eims) 1421 E1000_WRITE_REG(&adapter->hw, E1000_EIMS, que->eims); 1422 else 1423 igb_enable_intr(adapter); 1424} 1425 1426/* Deal with link in a sleepable context */ 1427static void 1428igb_handle_link(void *context, int pending) 1429{ 1430 struct adapter *adapter = context; 1431 1432 IGB_CORE_LOCK(adapter); 1433 igb_handle_link_locked(adapter); 1434 IGB_CORE_UNLOCK(adapter); 1435} 1436 1437static void 1438igb_handle_link_locked(struct adapter *adapter) 1439{ 1440 struct tx_ring *txr = adapter->tx_rings; 1441 struct ifnet *ifp = adapter->ifp; 1442 1443 IGB_CORE_LOCK_ASSERT(adapter); 1444 adapter->hw.mac.get_link_status = 1; 1445 igb_update_link_status(adapter); 1446 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) && adapter->link_active) { 1447 for (int i = 0; i < adapter->num_queues; i++, txr++) { 1448 IGB_TX_LOCK(txr); 1449#ifndef IGB_LEGACY_TX 1450 /* Process the stack queue only if not depleted */ 1451 if (((txr->queue_status & IGB_QUEUE_DEPLETED) == 0) && 1452 !drbr_empty(ifp, txr->br)) 1453 igb_mq_start_locked(ifp, txr); 1454#else 1455 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 1456 igb_start_locked(txr, ifp); 1457#endif 1458 IGB_TX_UNLOCK(txr); 1459 } 1460 } 1461} 1462 1463/********************************************************************* 1464 * 1465 * MSI/Legacy Deferred 1466 * Interrupt Service routine 1467 * 1468 *********************************************************************/ 1469static int 1470igb_irq_fast(void *arg) 1471{ 1472 struct adapter *adapter = arg; 1473 struct igb_queue *que = adapter->queues; 1474 u32 reg_icr; 1475 1476 1477 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR); 1478 1479 /* Hot eject? */ 1480 if (reg_icr == 0xffffffff) 1481 return FILTER_STRAY; 1482 1483 /* Definitely not our interrupt. */ 1484 if (reg_icr == 0x0) 1485 return FILTER_STRAY; 1486 1487 if ((reg_icr & E1000_ICR_INT_ASSERTED) == 0) 1488 return FILTER_STRAY; 1489 1490 /* 1491 * Mask interrupts until the taskqueue is finished running. This is 1492 * cheap, just assume that it is needed. This also works around the 1493 * MSI message reordering errata on certain systems. 1494 */ 1495 igb_disable_intr(adapter); 1496 taskqueue_enqueue(que->tq, &que->que_task); 1497 1498 /* Link status change */ 1499 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) 1500 taskqueue_enqueue(que->tq, &adapter->link_task); 1501 1502 if (reg_icr & E1000_ICR_RXO) 1503 adapter->rx_overruns++; 1504 return FILTER_HANDLED; 1505} 1506 1507#ifdef DEVICE_POLLING 1508#if __FreeBSD_version >= 800000 1509#define POLL_RETURN_COUNT(a) (a) 1510static int 1511#else 1512#define POLL_RETURN_COUNT(a) 1513static void 1514#endif 1515igb_poll(struct ifnet *ifp, enum poll_cmd cmd, int count) 1516{ 1517 struct adapter *adapter = ifp->if_softc; 1518 struct igb_queue *que; 1519 struct tx_ring *txr; 1520 u32 reg_icr, rx_done = 0; 1521 u32 loop = IGB_MAX_LOOP; 1522 bool more; 1523 1524 IGB_CORE_LOCK(adapter); 1525 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 1526 IGB_CORE_UNLOCK(adapter); 1527 return POLL_RETURN_COUNT(rx_done); 1528 } 1529 1530 if (cmd == POLL_AND_CHECK_STATUS) { 1531 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR); 1532 /* Link status change */ 1533 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) 1534 igb_handle_link_locked(adapter); 1535 1536 if (reg_icr & E1000_ICR_RXO) 1537 adapter->rx_overruns++; 1538 } 1539 IGB_CORE_UNLOCK(adapter); 1540 1541 for (int i = 0; i < adapter->num_queues; i++) { 1542 que = &adapter->queues[i]; 1543 txr = que->txr; 1544 1545 igb_rxeof(que, count, &rx_done); 1546 1547 IGB_TX_LOCK(txr); 1548 do { 1549 more = igb_txeof(txr); 1550 } while (loop-- && more); 1551#ifndef IGB_LEGACY_TX 1552 if (!drbr_empty(ifp, txr->br)) 1553 igb_mq_start_locked(ifp, txr); 1554#else 1555 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 1556 igb_start_locked(txr, ifp); 1557#endif 1558 IGB_TX_UNLOCK(txr); 1559 } 1560 1561 return POLL_RETURN_COUNT(rx_done); 1562} 1563#endif /* DEVICE_POLLING */ 1564 1565/********************************************************************* 1566 * 1567 * MSIX Que Interrupt Service routine 1568 * 1569 **********************************************************************/ 1570static void 1571igb_msix_que(void *arg) 1572{ 1573 struct igb_queue *que = arg; 1574 struct adapter *adapter = que->adapter; 1575 struct ifnet *ifp = adapter->ifp; 1576 struct tx_ring *txr = que->txr; 1577 struct rx_ring *rxr = que->rxr; 1578 u32 newitr = 0; 1579 bool more_rx; 1580 1581 /* Ignore spurious interrupts */ 1582 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 1583 return; 1584 1585 E1000_WRITE_REG(&adapter->hw, E1000_EIMC, que->eims); 1586 ++que->irqs; 1587 1588 IGB_TX_LOCK(txr); 1589 igb_txeof(txr); 1590#ifndef IGB_LEGACY_TX 1591 /* Process the stack queue only if not depleted */ 1592 if (((txr->queue_status & IGB_QUEUE_DEPLETED) == 0) && 1593 !drbr_empty(ifp, txr->br)) 1594 igb_mq_start_locked(ifp, txr); 1595#else 1596 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 1597 igb_start_locked(txr, ifp); 1598#endif 1599 IGB_TX_UNLOCK(txr); 1600 1601 more_rx = igb_rxeof(que, adapter->rx_process_limit, NULL); 1602 1603 if (adapter->enable_aim == FALSE) 1604 goto no_calc; 1605 /* 1606 ** Do Adaptive Interrupt Moderation: 1607 ** - Write out last calculated setting 1608 ** - Calculate based on average size over 1609 ** the last interval. 1610 */ 1611 if (que->eitr_setting) 1612 E1000_WRITE_REG(&adapter->hw, 1613 E1000_EITR(que->msix), que->eitr_setting); 1614 1615 que->eitr_setting = 0; 1616 1617 /* Idle, do nothing */ 1618 if ((txr->bytes == 0) && (rxr->bytes == 0)) 1619 goto no_calc; 1620 1621 /* Used half Default if sub-gig */ 1622 if (adapter->link_speed != 1000) 1623 newitr = IGB_DEFAULT_ITR / 2; 1624 else { 1625 if ((txr->bytes) && (txr->packets)) 1626 newitr = txr->bytes/txr->packets; 1627 if ((rxr->bytes) && (rxr->packets)) 1628 newitr = max(newitr, 1629 (rxr->bytes / rxr->packets)); 1630 newitr += 24; /* account for hardware frame, crc */ 1631 /* set an upper boundary */ 1632 newitr = min(newitr, 3000); 1633 /* Be nice to the mid range */ 1634 if ((newitr > 300) && (newitr < 1200)) 1635 newitr = (newitr / 3); 1636 else 1637 newitr = (newitr / 2); 1638 } 1639 newitr &= 0x7FFC; /* Mask invalid bits */ 1640 if (adapter->hw.mac.type == e1000_82575) 1641 newitr |= newitr << 16; 1642 else 1643 newitr |= E1000_EITR_CNT_IGNR; 1644 1645 /* save for next interrupt */ 1646 que->eitr_setting = newitr; 1647 1648 /* Reset state */ 1649 txr->bytes = 0; 1650 txr->packets = 0; 1651 rxr->bytes = 0; 1652 rxr->packets = 0; 1653 1654no_calc: 1655 /* Schedule a clean task if needed*/ 1656 if (more_rx) 1657 taskqueue_enqueue(que->tq, &que->que_task); 1658 else 1659 /* Reenable this interrupt */ 1660 E1000_WRITE_REG(&adapter->hw, E1000_EIMS, que->eims); 1661 return; 1662} 1663 1664 1665/********************************************************************* 1666 * 1667 * MSIX Link Interrupt Service routine 1668 * 1669 **********************************************************************/ 1670 1671static void 1672igb_msix_link(void *arg) 1673{ 1674 struct adapter *adapter = arg; 1675 u32 icr; 1676 1677 ++adapter->link_irq; 1678 icr = E1000_READ_REG(&adapter->hw, E1000_ICR); 1679 if (!(icr & E1000_ICR_LSC)) 1680 goto spurious; 1681 igb_handle_link(adapter, 0); 1682 1683spurious: 1684 /* Rearm */ 1685 E1000_WRITE_REG(&adapter->hw, E1000_IMS, E1000_IMS_LSC); 1686 E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask); 1687 return; 1688} 1689 1690 1691/********************************************************************* 1692 * 1693 * Media Ioctl callback 1694 * 1695 * This routine is called whenever the user queries the status of 1696 * the interface using ifconfig. 1697 * 1698 **********************************************************************/ 1699static void 1700igb_media_status(struct ifnet *ifp, struct ifmediareq *ifmr) 1701{ 1702 struct adapter *adapter = ifp->if_softc; 1703 1704 INIT_DEBUGOUT("igb_media_status: begin"); 1705 1706 IGB_CORE_LOCK(adapter); 1707 igb_update_link_status(adapter); 1708 1709 ifmr->ifm_status = IFM_AVALID; 1710 ifmr->ifm_active = IFM_ETHER; 1711 1712 if (!adapter->link_active) { 1713 IGB_CORE_UNLOCK(adapter); 1714 return; 1715 } 1716 1717 ifmr->ifm_status |= IFM_ACTIVE; 1718 1719 switch (adapter->link_speed) { 1720 case 10: 1721 ifmr->ifm_active |= IFM_10_T; 1722 break; 1723 case 100: 1724 /* 1725 ** Support for 100Mb SFP - these are Fiber 1726 ** but the media type appears as serdes 1727 */ 1728 if (adapter->hw.phy.media_type == 1729 e1000_media_type_internal_serdes) 1730 ifmr->ifm_active |= IFM_100_FX; 1731 else 1732 ifmr->ifm_active |= IFM_100_TX; 1733 break; 1734 case 1000: 1735 ifmr->ifm_active |= IFM_1000_T; 1736 break; 1737 } 1738 1739 if (adapter->link_duplex == FULL_DUPLEX) 1740 ifmr->ifm_active |= IFM_FDX; 1741 else 1742 ifmr->ifm_active |= IFM_HDX; 1743 1744 IGB_CORE_UNLOCK(adapter); 1745} 1746 1747/********************************************************************* 1748 * 1749 * Media Ioctl callback 1750 * 1751 * This routine is called when the user changes speed/duplex using 1752 * media/mediopt option with ifconfig. 1753 * 1754 **********************************************************************/ 1755static int 1756igb_media_change(struct ifnet *ifp) 1757{ 1758 struct adapter *adapter = ifp->if_softc; 1759 struct ifmedia *ifm = &adapter->media; 1760 1761 INIT_DEBUGOUT("igb_media_change: begin"); 1762 1763 if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) 1764 return (EINVAL); 1765 1766 IGB_CORE_LOCK(adapter); 1767 switch (IFM_SUBTYPE(ifm->ifm_media)) { 1768 case IFM_AUTO: 1769 adapter->hw.mac.autoneg = DO_AUTO_NEG; 1770 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT; 1771 break; 1772 case IFM_1000_LX: 1773 case IFM_1000_SX: 1774 case IFM_1000_T: 1775 adapter->hw.mac.autoneg = DO_AUTO_NEG; 1776 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL; 1777 break; 1778 case IFM_100_TX: 1779 adapter->hw.mac.autoneg = FALSE; 1780 adapter->hw.phy.autoneg_advertised = 0; 1781 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) 1782 adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL; 1783 else 1784 adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF; 1785 break; 1786 case IFM_10_T: 1787 adapter->hw.mac.autoneg = FALSE; 1788 adapter->hw.phy.autoneg_advertised = 0; 1789 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) 1790 adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL; 1791 else 1792 adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF; 1793 break; 1794 default: 1795 device_printf(adapter->dev, "Unsupported media type\n"); 1796 } 1797 1798 igb_init_locked(adapter); 1799 IGB_CORE_UNLOCK(adapter); 1800 1801 return (0); 1802} 1803 1804 1805/********************************************************************* 1806 * 1807 * This routine maps the mbufs to Advanced TX descriptors. 1808 * 1809 **********************************************************************/ 1810static int 1811igb_xmit(struct tx_ring *txr, struct mbuf **m_headp) 1812{ 1813 struct adapter *adapter = txr->adapter; 1814 bus_dma_segment_t segs[IGB_MAX_SCATTER]; 1815 bus_dmamap_t map; 1816 struct igb_tx_buffer *tx_buffer, *tx_buffer_mapped; 1817 union e1000_adv_tx_desc *txd = NULL; 1818 struct mbuf *m_head = *m_headp; 1819 struct ether_vlan_header *eh = NULL; 1820 struct ip *ip = NULL; 1821 struct tcphdr *th = NULL; 1822 u32 hdrlen, cmd_type_len, olinfo_status = 0; 1823 int ehdrlen, poff; 1824 int nsegs, i, first, last = 0; 1825 int error, do_tso, remap = 1; 1826 1827 /* Set basic descriptor constants */ 1828 cmd_type_len = E1000_ADVTXD_DTYP_DATA; 1829 cmd_type_len |= E1000_ADVTXD_DCMD_IFCS | E1000_ADVTXD_DCMD_DEXT; 1830 if (m_head->m_flags & M_VLANTAG) 1831 cmd_type_len |= E1000_ADVTXD_DCMD_VLE; 1832 1833retry: 1834 m_head = *m_headp; 1835 do_tso = ((m_head->m_pkthdr.csum_flags & CSUM_TSO) != 0); 1836 hdrlen = ehdrlen = poff = 0; 1837 1838 /* 1839 * Intel recommends entire IP/TCP header length reside in a single 1840 * buffer. If multiple descriptors are used to describe the IP and 1841 * TCP header, each descriptor should describe one or more 1842 * complete headers; descriptors referencing only parts of headers 1843 * are not supported. If all layer headers are not coalesced into 1844 * a single buffer, each buffer should not cross a 4KB boundary, 1845 * or be larger than the maximum read request size. 1846 * Controller also requires modifing IP/TCP header to make TSO work 1847 * so we firstly get a writable mbuf chain then coalesce ethernet/ 1848 * IP/TCP header into a single buffer to meet the requirement of 1849 * controller. This also simplifies IP/TCP/UDP checksum offloading 1850 * which also has similiar restrictions. 1851 */ 1852 if (do_tso || m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD) { 1853 if (do_tso || (m_head->m_next != NULL && 1854 m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD)) { 1855 if (M_WRITABLE(*m_headp) == 0) { 1856 m_head = m_dup(*m_headp, M_NOWAIT); 1857 m_freem(*m_headp); 1858 if (m_head == NULL) { 1859 *m_headp = NULL; 1860 return (ENOBUFS); 1861 } 1862 *m_headp = m_head; 1863 } 1864 } 1865 /* 1866 * Assume IPv4, we don't have TSO/checksum offload support 1867 * for IPv6 yet. 1868 */ 1869 ehdrlen = sizeof(struct ether_header); 1870 m_head = m_pullup(m_head, ehdrlen); 1871 if (m_head == NULL) { 1872 *m_headp = NULL; 1873 return (ENOBUFS); 1874 } 1875 eh = mtod(m_head, struct ether_vlan_header *); 1876 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { 1877 ehdrlen = sizeof(struct ether_vlan_header); 1878 m_head = m_pullup(m_head, ehdrlen); 1879 if (m_head == NULL) { 1880 *m_headp = NULL; 1881 return (ENOBUFS); 1882 } 1883 } 1884 m_head = m_pullup(m_head, ehdrlen + sizeof(struct ip)); 1885 if (m_head == NULL) { 1886 *m_headp = NULL; 1887 return (ENOBUFS); 1888 } 1889 ip = (struct ip *)(mtod(m_head, char *) + ehdrlen); 1890 poff = ehdrlen + (ip->ip_hl << 2); 1891 if (do_tso) { 1892 m_head = m_pullup(m_head, poff + sizeof(struct tcphdr)); 1893 if (m_head == NULL) { 1894 *m_headp = NULL; 1895 return (ENOBUFS); 1896 } 1897 /* 1898 * The pseudo TCP checksum does not include TCP payload 1899 * length so driver should recompute the checksum here 1900 * what hardware expect to see. This is adherence of 1901 * Microsoft's Large Send specification. 1902 */ 1903 th = (struct tcphdr *)(mtod(m_head, char *) + poff); 1904 th->th_sum = in_pseudo(ip->ip_src.s_addr, 1905 ip->ip_dst.s_addr, htons(IPPROTO_TCP)); 1906 /* Keep track of the full header length */ 1907 hdrlen = poff + (th->th_off << 2); 1908 } else if (m_head->m_pkthdr.csum_flags & CSUM_TCP) { 1909 m_head = m_pullup(m_head, poff + sizeof(struct tcphdr)); 1910 if (m_head == NULL) { 1911 *m_headp = NULL; 1912 return (ENOBUFS); 1913 } 1914 th = (struct tcphdr *)(mtod(m_head, char *) + poff); 1915 m_head = m_pullup(m_head, poff + (th->th_off << 2)); 1916 if (m_head == NULL) { 1917 *m_headp = NULL; 1918 return (ENOBUFS); 1919 } 1920 ip = (struct ip *)(mtod(m_head, char *) + ehdrlen); 1921 th = (struct tcphdr *)(mtod(m_head, char *) + poff); 1922 } else if (m_head->m_pkthdr.csum_flags & CSUM_UDP) { 1923 m_head = m_pullup(m_head, poff + sizeof(struct udphdr)); 1924 if (m_head == NULL) { 1925 *m_headp = NULL; 1926 return (ENOBUFS); 1927 } 1928 ip = (struct ip *)(mtod(m_head, char *) + ehdrlen); 1929 } 1930 *m_headp = m_head; 1931 } 1932 1933 /* 1934 * Map the packet for DMA 1935 * 1936 * Capture the first descriptor index, 1937 * this descriptor will have the index 1938 * of the EOP which is the only one that 1939 * now gets a DONE bit writeback. 1940 */ 1941 first = txr->next_avail_desc; 1942 tx_buffer = &txr->tx_buffers[first]; 1943 tx_buffer_mapped = tx_buffer; 1944 map = tx_buffer->map; 1945 1946 error = bus_dmamap_load_mbuf_sg(txr->txtag, map, 1947 *m_headp, segs, &nsegs, BUS_DMA_NOWAIT); 1948 1949 /* 1950 * There are two types of errors we can (try) to handle: 1951 * - EFBIG means the mbuf chain was too long and bus_dma ran 1952 * out of segments. Defragment the mbuf chain and try again. 1953 * - ENOMEM means bus_dma could not obtain enough bounce buffers 1954 * at this point in time. Defer sending and try again later. 1955 * All other errors, in particular EINVAL, are fatal and prevent the 1956 * mbuf chain from ever going through. Drop it and report error. 1957 */ 1958 if (error == EFBIG && remap) { 1959 struct mbuf *m; 1960 1961 m = m_defrag(*m_headp, M_NOWAIT); 1962 if (m == NULL) { 1963 adapter->mbuf_defrag_failed++; 1964 m_freem(*m_headp); 1965 *m_headp = NULL; 1966 return (ENOBUFS); 1967 } 1968 *m_headp = m; 1969 1970 /* Try it again, but only once */ 1971 remap = 0; 1972 goto retry; 1973 } else if (error == ENOMEM) { 1974 adapter->no_tx_dma_setup++; 1975 return (error); 1976 } else if (error != 0) { 1977 adapter->no_tx_dma_setup++; 1978 m_freem(*m_headp); 1979 *m_headp = NULL; 1980 return (error); 1981 } 1982 1983 /* 1984 ** Make sure we don't overrun the ring, 1985 ** we need nsegs descriptors and one for 1986 ** the context descriptor used for the 1987 ** offloads. 1988 */ 1989 if ((nsegs + 1) > (txr->tx_avail - 2)) { 1990 txr->no_desc_avail++; 1991 bus_dmamap_unload(txr->txtag, map); 1992 return (ENOBUFS); 1993 } 1994 m_head = *m_headp; 1995 1996 /* Do hardware assists: 1997 * Set up the context descriptor, used 1998 * when any hardware offload is done. 1999 * This includes CSUM, VLAN, and TSO. 2000 * It will use the first descriptor. 2001 */ 2002 2003 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 2004 if (igb_tso_setup(txr, m_head, ehdrlen, ip, th)) { 2005 cmd_type_len |= E1000_ADVTXD_DCMD_TSE; 2006 olinfo_status |= E1000_TXD_POPTS_IXSM << 8; 2007 olinfo_status |= E1000_TXD_POPTS_TXSM << 8; 2008 } else 2009 return (ENXIO); 2010 } else if (igb_tx_ctx_setup(txr, m_head)) 2011 olinfo_status |= E1000_TXD_POPTS_TXSM << 8; 2012 2013 /* Calculate payload length */ 2014 olinfo_status |= ((m_head->m_pkthdr.len - hdrlen) 2015 << E1000_ADVTXD_PAYLEN_SHIFT); 2016 2017 /* 82575 needs the queue index added */ 2018 if (adapter->hw.mac.type == e1000_82575) 2019 olinfo_status |= txr->me << 4; 2020 2021 /* Set up our transmit descriptors */ 2022 i = txr->next_avail_desc; 2023 for (int j = 0; j < nsegs; j++) { 2024 bus_size_t seg_len; 2025 bus_addr_t seg_addr; 2026 2027 tx_buffer = &txr->tx_buffers[i]; 2028 txd = (union e1000_adv_tx_desc *)&txr->tx_base[i]; 2029 seg_addr = segs[j].ds_addr; 2030 seg_len = segs[j].ds_len; 2031 2032 txd->read.buffer_addr = htole64(seg_addr); 2033 txd->read.cmd_type_len = htole32(cmd_type_len | seg_len); 2034 txd->read.olinfo_status = htole32(olinfo_status); 2035 last = i; 2036 if (++i == adapter->num_tx_desc) 2037 i = 0; 2038 tx_buffer->m_head = NULL; 2039 tx_buffer->next_eop = -1; 2040 } 2041 2042 txr->next_avail_desc = i; 2043 txr->tx_avail -= nsegs; 2044 tx_buffer->m_head = m_head; 2045 2046 /* 2047 ** Here we swap the map so the last descriptor, 2048 ** which gets the completion interrupt has the 2049 ** real map, and the first descriptor gets the 2050 ** unused map from this descriptor. 2051 */ 2052 tx_buffer_mapped->map = tx_buffer->map; 2053 tx_buffer->map = map; 2054 bus_dmamap_sync(txr->txtag, map, BUS_DMASYNC_PREWRITE); 2055 2056 /* 2057 * Last Descriptor of Packet 2058 * needs End Of Packet (EOP) 2059 * and Report Status (RS) 2060 */ 2061 txd->read.cmd_type_len |= 2062 htole32(E1000_ADVTXD_DCMD_EOP | E1000_ADVTXD_DCMD_RS); 2063 /* 2064 * Keep track in the first buffer which 2065 * descriptor will be written back 2066 */ 2067 tx_buffer = &txr->tx_buffers[first]; 2068 tx_buffer->next_eop = last; 2069 /* Update the watchdog time early and often */ 2070 txr->watchdog_time = ticks; 2071 2072 /* 2073 * Advance the Transmit Descriptor Tail (TDT), this tells the E1000 2074 * that this frame is available to transmit. 2075 */ 2076 bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, 2077 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 2078 E1000_WRITE_REG(&adapter->hw, E1000_TDT(txr->me), i); 2079 ++txr->tx_packets; 2080 2081 return (0); 2082} 2083static void 2084igb_set_promisc(struct adapter *adapter) 2085{ 2086 struct ifnet *ifp = adapter->ifp; 2087 struct e1000_hw *hw = &adapter->hw; 2088 u32 reg; 2089 2090 if (adapter->vf_ifp) { 2091 e1000_promisc_set_vf(hw, e1000_promisc_enabled); 2092 return; 2093 } 2094 2095 reg = E1000_READ_REG(hw, E1000_RCTL); 2096 if (ifp->if_flags & IFF_PROMISC) { 2097 reg |= (E1000_RCTL_UPE | E1000_RCTL_MPE); 2098 E1000_WRITE_REG(hw, E1000_RCTL, reg); 2099 } else if (ifp->if_flags & IFF_ALLMULTI) { 2100 reg |= E1000_RCTL_MPE; 2101 reg &= ~E1000_RCTL_UPE; 2102 E1000_WRITE_REG(hw, E1000_RCTL, reg); 2103 } 2104} 2105 2106static void 2107igb_disable_promisc(struct adapter *adapter) 2108{ 2109 struct e1000_hw *hw = &adapter->hw; 2110 struct ifnet *ifp = adapter->ifp; 2111 u32 reg; 2112 int mcnt = 0; 2113 2114 if (adapter->vf_ifp) { 2115 e1000_promisc_set_vf(hw, e1000_promisc_disabled); 2116 return; 2117 } 2118 reg = E1000_READ_REG(hw, E1000_RCTL); 2119 reg &= (~E1000_RCTL_UPE); 2120 if (ifp->if_flags & IFF_ALLMULTI) 2121 mcnt = MAX_NUM_MULTICAST_ADDRESSES; 2122 else { 2123 struct ifmultiaddr *ifma; 2124#if __FreeBSD_version < 800000 2125 IF_ADDR_LOCK(ifp); 2126#else 2127 if_maddr_rlock(ifp); 2128#endif 2129 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2130 if (ifma->ifma_addr->sa_family != AF_LINK) 2131 continue; 2132 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES) 2133 break; 2134 mcnt++; 2135 } 2136#if __FreeBSD_version < 800000 2137 IF_ADDR_UNLOCK(ifp); 2138#else 2139 if_maddr_runlock(ifp); 2140#endif 2141 } 2142 /* Don't disable if in MAX groups */ 2143 if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) 2144 reg &= (~E1000_RCTL_MPE); 2145 E1000_WRITE_REG(hw, E1000_RCTL, reg); 2146} 2147 2148 2149/********************************************************************* 2150 * Multicast Update 2151 * 2152 * This routine is called whenever multicast address list is updated. 2153 * 2154 **********************************************************************/ 2155 2156static void 2157igb_set_multi(struct adapter *adapter) 2158{ 2159 struct ifnet *ifp = adapter->ifp; 2160 struct ifmultiaddr *ifma; 2161 u32 reg_rctl = 0; 2162 u8 *mta; 2163 2164 int mcnt = 0; 2165 2166 IOCTL_DEBUGOUT("igb_set_multi: begin"); 2167 2168 mta = adapter->mta; 2169 bzero(mta, sizeof(uint8_t) * ETH_ADDR_LEN * 2170 MAX_NUM_MULTICAST_ADDRESSES); 2171 2172#if __FreeBSD_version < 800000 2173 IF_ADDR_LOCK(ifp); 2174#else 2175 if_maddr_rlock(ifp); 2176#endif 2177 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2178 if (ifma->ifma_addr->sa_family != AF_LINK) 2179 continue; 2180 2181 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES) 2182 break; 2183 2184 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 2185 &mta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 2186 mcnt++; 2187 } 2188#if __FreeBSD_version < 800000 2189 IF_ADDR_UNLOCK(ifp); 2190#else 2191 if_maddr_runlock(ifp); 2192#endif 2193 2194 if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) { 2195 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL); 2196 reg_rctl |= E1000_RCTL_MPE; 2197 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl); 2198 } else 2199 e1000_update_mc_addr_list(&adapter->hw, mta, mcnt); 2200} 2201 2202 2203/********************************************************************* 2204 * Timer routine: 2205 * This routine checks for link status, 2206 * updates statistics, and does the watchdog. 2207 * 2208 **********************************************************************/ 2209 2210static void 2211igb_local_timer(void *arg) 2212{ 2213 struct adapter *adapter = arg; 2214 device_t dev = adapter->dev; 2215 struct ifnet *ifp = adapter->ifp; 2216 struct tx_ring *txr = adapter->tx_rings; 2217 struct igb_queue *que = adapter->queues; 2218 int hung = 0, busy = 0; 2219 2220 2221 IGB_CORE_LOCK_ASSERT(adapter); 2222 2223 igb_update_link_status(adapter); 2224 igb_update_stats_counters(adapter); 2225 2226 /* 2227 ** Check the TX queues status 2228 ** - central locked handling of OACTIVE 2229 ** - watchdog only if all queues show hung 2230 */ 2231 for (int i = 0; i < adapter->num_queues; i++, que++, txr++) { 2232 if ((txr->queue_status & IGB_QUEUE_HUNG) && 2233 (adapter->pause_frames == 0)) 2234 ++hung; 2235 if (txr->queue_status & IGB_QUEUE_DEPLETED) 2236 ++busy; 2237 if ((txr->queue_status & IGB_QUEUE_IDLE) == 0) 2238 taskqueue_enqueue(que->tq, &que->que_task); 2239 } 2240 if (hung == adapter->num_queues) 2241 goto timeout; 2242 if (busy == adapter->num_queues) 2243 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 2244 else if ((ifp->if_drv_flags & IFF_DRV_OACTIVE) && 2245 (busy < adapter->num_queues)) 2246 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 2247 2248 adapter->pause_frames = 0; 2249 callout_reset(&adapter->timer, hz, igb_local_timer, adapter); 2250#ifndef DEVICE_POLLING 2251 /* Schedule all queue interrupts - deadlock protection */ 2252 E1000_WRITE_REG(&adapter->hw, E1000_EICS, adapter->que_mask); 2253#endif 2254 return; 2255 2256timeout: 2257 device_printf(adapter->dev, "Watchdog timeout -- resetting\n"); 2258 device_printf(dev,"Queue(%d) tdh = %d, hw tdt = %d\n", txr->me, 2259 E1000_READ_REG(&adapter->hw, E1000_TDH(txr->me)), 2260 E1000_READ_REG(&adapter->hw, E1000_TDT(txr->me))); 2261 device_printf(dev,"TX(%d) desc avail = %d," 2262 "Next TX to Clean = %d\n", 2263 txr->me, txr->tx_avail, txr->next_to_clean); 2264 adapter->ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 2265 adapter->watchdog_events++; 2266 igb_init_locked(adapter); 2267} 2268 2269static void 2270igb_update_link_status(struct adapter *adapter) 2271{ 2272 struct e1000_hw *hw = &adapter->hw; 2273 struct e1000_fc_info *fc = &hw->fc; 2274 struct ifnet *ifp = adapter->ifp; 2275 device_t dev = adapter->dev; 2276 struct tx_ring *txr = adapter->tx_rings; 2277 u32 link_check, thstat, ctrl; 2278 char *flowctl = NULL; 2279 2280 link_check = thstat = ctrl = 0; 2281 2282 /* Get the cached link value or read for real */ 2283 switch (hw->phy.media_type) { 2284 case e1000_media_type_copper: 2285 if (hw->mac.get_link_status) { 2286 /* Do the work to read phy */ 2287 e1000_check_for_link(hw); 2288 link_check = !hw->mac.get_link_status; 2289 } else 2290 link_check = TRUE; 2291 break; 2292 case e1000_media_type_fiber: 2293 e1000_check_for_link(hw); 2294 link_check = (E1000_READ_REG(hw, E1000_STATUS) & 2295 E1000_STATUS_LU); 2296 break; 2297 case e1000_media_type_internal_serdes: 2298 e1000_check_for_link(hw); 2299 link_check = adapter->hw.mac.serdes_has_link; 2300 break; 2301 /* VF device is type_unknown */ 2302 case e1000_media_type_unknown: 2303 e1000_check_for_link(hw); 2304 link_check = !hw->mac.get_link_status; 2305 /* Fall thru */ 2306 default: 2307 break; 2308 } 2309 2310 /* Check for thermal downshift or shutdown */ 2311 if (hw->mac.type == e1000_i350) { 2312 thstat = E1000_READ_REG(hw, E1000_THSTAT); 2313 ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT); 2314 } 2315 2316 /* Get the flow control for display */ 2317 switch (fc->current_mode) { 2318 case e1000_fc_rx_pause: 2319 flowctl = "RX"; 2320 break; 2321 case e1000_fc_tx_pause: 2322 flowctl = "TX"; 2323 break; 2324 case e1000_fc_full: 2325 flowctl = "Full"; 2326 break; 2327 case e1000_fc_none: 2328 default: 2329 flowctl = "None"; 2330 break; 2331 } 2332 2333 /* Now we check if a transition has happened */ 2334 if (link_check && (adapter->link_active == 0)) { 2335 e1000_get_speed_and_duplex(&adapter->hw, 2336 &adapter->link_speed, &adapter->link_duplex); 2337 if (bootverbose) 2338 device_printf(dev, "Link is up %d Mbps %s," 2339 " Flow Control: %s\n", 2340 adapter->link_speed, 2341 ((adapter->link_duplex == FULL_DUPLEX) ? 2342 "Full Duplex" : "Half Duplex"), flowctl); 2343 adapter->link_active = 1; 2344 ifp->if_baudrate = adapter->link_speed * 1000000; 2345 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) && 2346 (thstat & E1000_THSTAT_LINK_THROTTLE)) 2347 device_printf(dev, "Link: thermal downshift\n"); 2348 /* This can sleep */ 2349 if_link_state_change(ifp, LINK_STATE_UP); 2350 } else if (!link_check && (adapter->link_active == 1)) { 2351 ifp->if_baudrate = adapter->link_speed = 0; 2352 adapter->link_duplex = 0; 2353 if (bootverbose) 2354 device_printf(dev, "Link is Down\n"); 2355 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) && 2356 (thstat & E1000_THSTAT_PWR_DOWN)) 2357 device_printf(dev, "Link: thermal shutdown\n"); 2358 adapter->link_active = 0; 2359 /* This can sleep */ 2360 if_link_state_change(ifp, LINK_STATE_DOWN); 2361 /* Reset queue state */ 2362 for (int i = 0; i < adapter->num_queues; i++, txr++) 2363 txr->queue_status = IGB_QUEUE_IDLE; 2364 } 2365} 2366 2367/********************************************************************* 2368 * 2369 * This routine disables all traffic on the adapter by issuing a 2370 * global reset on the MAC and deallocates TX/RX buffers. 2371 * 2372 **********************************************************************/ 2373 2374static void 2375igb_stop(void *arg) 2376{ 2377 struct adapter *adapter = arg; 2378 struct ifnet *ifp = adapter->ifp; 2379 struct tx_ring *txr = adapter->tx_rings; 2380 2381 IGB_CORE_LOCK_ASSERT(adapter); 2382 2383 INIT_DEBUGOUT("igb_stop: begin"); 2384 2385 igb_disable_intr(adapter); 2386 2387 callout_stop(&adapter->timer); 2388 2389 /* Tell the stack that the interface is no longer active */ 2390 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 2391 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 2392 2393 /* Disarm watchdog timer. */ 2394 for (int i = 0; i < adapter->num_queues; i++, txr++) { 2395 IGB_TX_LOCK(txr); 2396 txr->queue_status = IGB_QUEUE_IDLE; 2397 IGB_TX_UNLOCK(txr); 2398 } 2399 2400 e1000_reset_hw(&adapter->hw); 2401 E1000_WRITE_REG(&adapter->hw, E1000_WUC, 0); 2402 2403 e1000_led_off(&adapter->hw); 2404 e1000_cleanup_led(&adapter->hw); 2405} 2406 2407 2408/********************************************************************* 2409 * 2410 * Determine hardware revision. 2411 * 2412 **********************************************************************/ 2413static void 2414igb_identify_hardware(struct adapter *adapter) 2415{ 2416 device_t dev = adapter->dev; 2417 2418 /* Make sure our PCI config space has the necessary stuff set */ 2419 pci_enable_busmaster(dev); 2420 adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2); 2421 2422 /* Save off the information about this board */ 2423 adapter->hw.vendor_id = pci_get_vendor(dev); 2424 adapter->hw.device_id = pci_get_device(dev); 2425 adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1); 2426 adapter->hw.subsystem_vendor_id = 2427 pci_read_config(dev, PCIR_SUBVEND_0, 2); 2428 adapter->hw.subsystem_device_id = 2429 pci_read_config(dev, PCIR_SUBDEV_0, 2); 2430 2431 /* Set MAC type early for PCI setup */ 2432 e1000_set_mac_type(&adapter->hw); 2433 2434 /* Are we a VF device? */ 2435 if ((adapter->hw.mac.type == e1000_vfadapt) || 2436 (adapter->hw.mac.type == e1000_vfadapt_i350)) 2437 adapter->vf_ifp = 1; 2438 else 2439 adapter->vf_ifp = 0; 2440} 2441 2442static int 2443igb_allocate_pci_resources(struct adapter *adapter) 2444{ 2445 device_t dev = adapter->dev; 2446 int rid; 2447 2448 rid = PCIR_BAR(0); 2449 adapter->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, 2450 &rid, RF_ACTIVE); 2451 if (adapter->pci_mem == NULL) { 2452 device_printf(dev, "Unable to allocate bus resource: memory\n"); 2453 return (ENXIO); 2454 } 2455 adapter->osdep.mem_bus_space_tag = 2456 rman_get_bustag(adapter->pci_mem); 2457 adapter->osdep.mem_bus_space_handle = 2458 rman_get_bushandle(adapter->pci_mem); 2459 adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle; 2460 2461 adapter->num_queues = 1; /* Defaults for Legacy or MSI */ 2462 2463 /* This will setup either MSI/X or MSI */ 2464 adapter->msix = igb_setup_msix(adapter); 2465 adapter->hw.back = &adapter->osdep; 2466 2467 return (0); 2468} 2469 2470/********************************************************************* 2471 * 2472 * Setup the Legacy or MSI Interrupt handler 2473 * 2474 **********************************************************************/ 2475static int 2476igb_allocate_legacy(struct adapter *adapter) 2477{ 2478 device_t dev = adapter->dev; 2479 struct igb_queue *que = adapter->queues; 2480 int error, rid = 0; 2481 2482 /* Turn off all interrupts */ 2483 E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff); 2484 2485 /* MSI RID is 1 */ 2486 if (adapter->msix == 1) 2487 rid = 1; 2488 2489 /* We allocate a single interrupt resource */ 2490 adapter->res = bus_alloc_resource_any(dev, 2491 SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE); 2492 if (adapter->res == NULL) { 2493 device_printf(dev, "Unable to allocate bus resource: " 2494 "interrupt\n"); 2495 return (ENXIO); 2496 } 2497 2498#ifndef IGB_LEGACY_TX 2499 TASK_INIT(&que->txr->txq_task, 0, igb_deferred_mq_start, que->txr); 2500#endif 2501 2502 /* 2503 * Try allocating a fast interrupt and the associated deferred 2504 * processing contexts. 2505 */ 2506 TASK_INIT(&que->que_task, 0, igb_handle_que, que); 2507 /* Make tasklet for deferred link handling */ 2508 TASK_INIT(&adapter->link_task, 0, igb_handle_link, adapter); 2509 que->tq = taskqueue_create_fast("igb_taskq", M_NOWAIT, 2510 taskqueue_thread_enqueue, &que->tq); 2511 taskqueue_start_threads(&que->tq, 1, PI_NET, "%s taskq", 2512 device_get_nameunit(adapter->dev)); 2513 if ((error = bus_setup_intr(dev, adapter->res, 2514 INTR_TYPE_NET | INTR_MPSAFE, igb_irq_fast, NULL, 2515 adapter, &adapter->tag)) != 0) { 2516 device_printf(dev, "Failed to register fast interrupt " 2517 "handler: %d\n", error); 2518 taskqueue_free(que->tq); 2519 que->tq = NULL; 2520 return (error); 2521 } 2522 2523 return (0); 2524} 2525 2526 2527/********************************************************************* 2528 * 2529 * Setup the MSIX Queue Interrupt handlers: 2530 * 2531 **********************************************************************/ 2532static int 2533igb_allocate_msix(struct adapter *adapter) 2534{ 2535 device_t dev = adapter->dev; 2536 struct igb_queue *que = adapter->queues; 2537 int error, rid, vector = 0; 2538 2539 /* Be sure to start with all interrupts disabled */ 2540 E1000_WRITE_REG(&adapter->hw, E1000_IMC, ~0); 2541 E1000_WRITE_FLUSH(&adapter->hw); 2542 2543 for (int i = 0; i < adapter->num_queues; i++, vector++, que++) { 2544 rid = vector +1; 2545 que->res = bus_alloc_resource_any(dev, 2546 SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE); 2547 if (que->res == NULL) { 2548 device_printf(dev, 2549 "Unable to allocate bus resource: " 2550 "MSIX Queue Interrupt\n"); 2551 return (ENXIO); 2552 } 2553 error = bus_setup_intr(dev, que->res, 2554 INTR_TYPE_NET | INTR_MPSAFE, NULL, 2555 igb_msix_que, que, &que->tag); 2556 if (error) { 2557 que->res = NULL; 2558 device_printf(dev, "Failed to register Queue handler"); 2559 return (error); 2560 } 2561#if __FreeBSD_version >= 800504 2562 bus_describe_intr(dev, que->res, que->tag, "que %d", i); 2563#endif 2564 que->msix = vector; 2565 if (adapter->hw.mac.type == e1000_82575) 2566 que->eims = E1000_EICR_TX_QUEUE0 << i; 2567 else 2568 que->eims = 1 << vector; 2569 /* 2570 ** Bind the msix vector, and thus the 2571 ** rings to the corresponding cpu. 2572 */ 2573 if (adapter->num_queues > 1) { 2574 if (igb_last_bind_cpu < 0) 2575 igb_last_bind_cpu = CPU_FIRST(); 2576 bus_bind_intr(dev, que->res, igb_last_bind_cpu); 2577 device_printf(dev, 2578 "Bound queue %d to cpu %d\n", 2579 i,igb_last_bind_cpu); 2580 igb_last_bind_cpu = CPU_NEXT(igb_last_bind_cpu); 2581 } 2582#ifndef IGB_LEGACY_TX 2583 TASK_INIT(&que->txr->txq_task, 0, igb_deferred_mq_start, 2584 que->txr); 2585#endif 2586 /* Make tasklet for deferred handling */ 2587 TASK_INIT(&que->que_task, 0, igb_handle_que, que); 2588 que->tq = taskqueue_create("igb_que", M_NOWAIT, 2589 taskqueue_thread_enqueue, &que->tq); 2590 taskqueue_start_threads(&que->tq, 1, PI_NET, "%s que", 2591 device_get_nameunit(adapter->dev)); 2592 } 2593 2594 /* And Link */ 2595 rid = vector + 1; 2596 adapter->res = bus_alloc_resource_any(dev, 2597 SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE); 2598 if (adapter->res == NULL) { 2599 device_printf(dev, 2600 "Unable to allocate bus resource: " 2601 "MSIX Link Interrupt\n"); 2602 return (ENXIO); 2603 } 2604 if ((error = bus_setup_intr(dev, adapter->res, 2605 INTR_TYPE_NET | INTR_MPSAFE, NULL, 2606 igb_msix_link, adapter, &adapter->tag)) != 0) { 2607 device_printf(dev, "Failed to register Link handler"); 2608 return (error); 2609 } 2610#if __FreeBSD_version >= 800504 2611 bus_describe_intr(dev, adapter->res, adapter->tag, "link"); 2612#endif 2613 adapter->linkvec = vector; 2614 2615 return (0); 2616} 2617 2618 2619static void 2620igb_configure_queues(struct adapter *adapter) 2621{ 2622 struct e1000_hw *hw = &adapter->hw; 2623 struct igb_queue *que; 2624 u32 tmp, ivar = 0, newitr = 0; 2625 2626 /* First turn on RSS capability */ 2627 if (adapter->hw.mac.type != e1000_82575) 2628 E1000_WRITE_REG(hw, E1000_GPIE, 2629 E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME | 2630 E1000_GPIE_PBA | E1000_GPIE_NSICR); 2631 2632 /* Turn on MSIX */ 2633 switch (adapter->hw.mac.type) { 2634 case e1000_82580: 2635 case e1000_i350: 2636 case e1000_i210: 2637 case e1000_i211: 2638 case e1000_vfadapt: 2639 case e1000_vfadapt_i350: 2640 /* RX entries */ 2641 for (int i = 0; i < adapter->num_queues; i++) { 2642 u32 index = i >> 1; 2643 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 2644 que = &adapter->queues[i]; 2645 if (i & 1) { 2646 ivar &= 0xFF00FFFF; 2647 ivar |= (que->msix | E1000_IVAR_VALID) << 16; 2648 } else { 2649 ivar &= 0xFFFFFF00; 2650 ivar |= que->msix | E1000_IVAR_VALID; 2651 } 2652 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 2653 } 2654 /* TX entries */ 2655 for (int i = 0; i < adapter->num_queues; i++) { 2656 u32 index = i >> 1; 2657 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 2658 que = &adapter->queues[i]; 2659 if (i & 1) { 2660 ivar &= 0x00FFFFFF; 2661 ivar |= (que->msix | E1000_IVAR_VALID) << 24; 2662 } else { 2663 ivar &= 0xFFFF00FF; 2664 ivar |= (que->msix | E1000_IVAR_VALID) << 8; 2665 } 2666 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 2667 adapter->que_mask |= que->eims; 2668 } 2669 2670 /* And for the link interrupt */ 2671 ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8; 2672 adapter->link_mask = 1 << adapter->linkvec; 2673 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar); 2674 break; 2675 case e1000_82576: 2676 /* RX entries */ 2677 for (int i = 0; i < adapter->num_queues; i++) { 2678 u32 index = i & 0x7; /* Each IVAR has two entries */ 2679 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 2680 que = &adapter->queues[i]; 2681 if (i < 8) { 2682 ivar &= 0xFFFFFF00; 2683 ivar |= que->msix | E1000_IVAR_VALID; 2684 } else { 2685 ivar &= 0xFF00FFFF; 2686 ivar |= (que->msix | E1000_IVAR_VALID) << 16; 2687 } 2688 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 2689 adapter->que_mask |= que->eims; 2690 } 2691 /* TX entries */ 2692 for (int i = 0; i < adapter->num_queues; i++) { 2693 u32 index = i & 0x7; /* Each IVAR has two entries */ 2694 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 2695 que = &adapter->queues[i]; 2696 if (i < 8) { 2697 ivar &= 0xFFFF00FF; 2698 ivar |= (que->msix | E1000_IVAR_VALID) << 8; 2699 } else { 2700 ivar &= 0x00FFFFFF; 2701 ivar |= (que->msix | E1000_IVAR_VALID) << 24; 2702 } 2703 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 2704 adapter->que_mask |= que->eims; 2705 } 2706 2707 /* And for the link interrupt */ 2708 ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8; 2709 adapter->link_mask = 1 << adapter->linkvec; 2710 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar); 2711 break; 2712 2713 case e1000_82575: 2714 /* enable MSI-X support*/ 2715 tmp = E1000_READ_REG(hw, E1000_CTRL_EXT); 2716 tmp |= E1000_CTRL_EXT_PBA_CLR; 2717 /* Auto-Mask interrupts upon ICR read. */ 2718 tmp |= E1000_CTRL_EXT_EIAME; 2719 tmp |= E1000_CTRL_EXT_IRCA; 2720 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp); 2721 2722 /* Queues */ 2723 for (int i = 0; i < adapter->num_queues; i++) { 2724 que = &adapter->queues[i]; 2725 tmp = E1000_EICR_RX_QUEUE0 << i; 2726 tmp |= E1000_EICR_TX_QUEUE0 << i; 2727 que->eims = tmp; 2728 E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 2729 i, que->eims); 2730 adapter->que_mask |= que->eims; 2731 } 2732 2733 /* Link */ 2734 E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec), 2735 E1000_EIMS_OTHER); 2736 adapter->link_mask |= E1000_EIMS_OTHER; 2737 default: 2738 break; 2739 } 2740 2741 /* Set the starting interrupt rate */ 2742 if (igb_max_interrupt_rate > 0) 2743 newitr = (4000000 / igb_max_interrupt_rate) & 0x7FFC; 2744 2745 if (hw->mac.type == e1000_82575) 2746 newitr |= newitr << 16; 2747 else 2748 newitr |= E1000_EITR_CNT_IGNR; 2749 2750 for (int i = 0; i < adapter->num_queues; i++) { 2751 que = &adapter->queues[i]; 2752 E1000_WRITE_REG(hw, E1000_EITR(que->msix), newitr); 2753 } 2754 2755 return; 2756} 2757 2758 2759static void 2760igb_free_pci_resources(struct adapter *adapter) 2761{ 2762 struct igb_queue *que = adapter->queues; 2763 device_t dev = adapter->dev; 2764 int rid; 2765 2766 /* 2767 ** There is a slight possibility of a failure mode 2768 ** in attach that will result in entering this function 2769 ** before interrupt resources have been initialized, and 2770 ** in that case we do not want to execute the loops below 2771 ** We can detect this reliably by the state of the adapter 2772 ** res pointer. 2773 */ 2774 if (adapter->res == NULL) 2775 goto mem; 2776 2777 /* 2778 * First release all the interrupt resources: 2779 */ 2780 for (int i = 0; i < adapter->num_queues; i++, que++) { 2781 rid = que->msix + 1; 2782 if (que->tag != NULL) { 2783 bus_teardown_intr(dev, que->res, que->tag); 2784 que->tag = NULL; 2785 } 2786 if (que->res != NULL) 2787 bus_release_resource(dev, 2788 SYS_RES_IRQ, rid, que->res); 2789 } 2790 2791 /* Clean the Legacy or Link interrupt last */ 2792 if (adapter->linkvec) /* we are doing MSIX */ 2793 rid = adapter->linkvec + 1; 2794 else 2795 (adapter->msix != 0) ? (rid = 1):(rid = 0); 2796 2797 que = adapter->queues; 2798 if (adapter->tag != NULL) { 2799 taskqueue_drain(que->tq, &adapter->link_task); 2800 bus_teardown_intr(dev, adapter->res, adapter->tag); 2801 adapter->tag = NULL; 2802 } 2803 if (adapter->res != NULL) 2804 bus_release_resource(dev, SYS_RES_IRQ, rid, adapter->res); 2805 2806 for (int i = 0; i < adapter->num_queues; i++, que++) { 2807 if (que->tq != NULL) { 2808#ifndef IGB_LEGACY_TX 2809 taskqueue_drain(que->tq, &que->txr->txq_task); 2810#endif 2811 taskqueue_drain(que->tq, &que->que_task); 2812 taskqueue_free(que->tq); 2813 } 2814 } 2815mem: 2816 if (adapter->msix) 2817 pci_release_msi(dev); 2818 2819 if (adapter->msix_mem != NULL) 2820 bus_release_resource(dev, SYS_RES_MEMORY, 2821 PCIR_BAR(IGB_MSIX_BAR), adapter->msix_mem); 2822 2823 if (adapter->pci_mem != NULL) 2824 bus_release_resource(dev, SYS_RES_MEMORY, 2825 PCIR_BAR(0), adapter->pci_mem); 2826 2827} 2828 2829/* 2830 * Setup Either MSI/X or MSI 2831 */ 2832static int 2833igb_setup_msix(struct adapter *adapter) 2834{ 2835 device_t dev = adapter->dev; 2836 int rid, want, queues, msgs, maxqueues; 2837 2838 /* tuneable override */ 2839 if (igb_enable_msix == 0) 2840 goto msi; 2841 2842 /* First try MSI/X */ 2843 msgs = pci_msix_count(dev); 2844 if (msgs == 0) 2845 goto msi; 2846 rid = PCIR_BAR(IGB_MSIX_BAR); 2847 adapter->msix_mem = bus_alloc_resource_any(dev, 2848 SYS_RES_MEMORY, &rid, RF_ACTIVE); 2849 if (adapter->msix_mem == NULL) { 2850 /* May not be enabled */ 2851 device_printf(adapter->dev, 2852 "Unable to map MSIX table \n"); 2853 goto msi; 2854 } 2855 2856 /* Figure out a reasonable auto config value */ 2857 queues = (mp_ncpus > (msgs-1)) ? (msgs-1) : mp_ncpus; 2858 2859 /* Manual override */ 2860 if (igb_num_queues != 0) 2861 queues = igb_num_queues; 2862 2863 /* Sanity check based on HW */ 2864 switch (adapter->hw.mac.type) { 2865 case e1000_82575: 2866 maxqueues = 4; 2867 break; 2868 case e1000_82576: 2869 case e1000_82580: 2870 case e1000_i350: 2871 maxqueues = 8; 2872 break; 2873 case e1000_i210: 2874 maxqueues = 4; 2875 break; 2876 case e1000_i211: 2877 maxqueues = 2; 2878 break; 2879 default: /* VF interfaces */ 2880 maxqueues = 1; 2881 break; 2882 } 2883 if (queues > maxqueues) 2884 queues = maxqueues; 2885 2886 /* 2887 ** One vector (RX/TX pair) per queue 2888 ** plus an additional for Link interrupt 2889 */ 2890 want = queues + 1; 2891 if (msgs >= want) 2892 msgs = want; 2893 else { 2894 device_printf(adapter->dev, 2895 "MSIX Configuration Problem, " 2896 "%d vectors configured, but %d queues wanted!\n", 2897 msgs, want); 2898 goto msi; 2899 } 2900 if ((pci_alloc_msix(dev, &msgs) == 0) && (msgs == want)) { 2901 device_printf(adapter->dev, 2902 "Using MSIX interrupts with %d vectors\n", msgs); 2903 adapter->num_queues = queues; 2904 return (msgs); 2905 } 2906 /* 2907 ** If MSIX alloc failed or provided us with 2908 ** less than needed, free and fall through to MSI 2909 */ 2910 pci_release_msi(dev); 2911 2912msi: 2913 if (adapter->msix_mem != NULL) { 2914 bus_release_resource(dev, SYS_RES_MEMORY, 2915 PCIR_BAR(IGB_MSIX_BAR), adapter->msix_mem); 2916 adapter->msix_mem = NULL; 2917 } 2918 msgs = 1; 2919 if (pci_alloc_msi(dev, &msgs) == 0) { 2920 device_printf(adapter->dev," Using an MSI interrupt\n"); 2921 return (msgs); 2922 } 2923 device_printf(adapter->dev," Using a Legacy interrupt\n"); 2924 return (0); 2925} 2926 2927/********************************************************************* 2928 * 2929 * Set up an fresh starting state 2930 * 2931 **********************************************************************/ 2932static void 2933igb_reset(struct adapter *adapter) 2934{ 2935 device_t dev = adapter->dev; 2936 struct e1000_hw *hw = &adapter->hw; 2937 struct e1000_fc_info *fc = &hw->fc; 2938 struct ifnet *ifp = adapter->ifp; 2939 u32 pba = 0; 2940 u16 hwm; 2941 2942 INIT_DEBUGOUT("igb_reset: begin"); 2943 2944 /* Let the firmware know the OS is in control */ 2945 igb_get_hw_control(adapter); 2946 2947 /* 2948 * Packet Buffer Allocation (PBA) 2949 * Writing PBA sets the receive portion of the buffer 2950 * the remainder is used for the transmit buffer. 2951 */ 2952 switch (hw->mac.type) { 2953 case e1000_82575: 2954 pba = E1000_PBA_32K; 2955 break; 2956 case e1000_82576: 2957 case e1000_vfadapt: 2958 pba = E1000_READ_REG(hw, E1000_RXPBS); 2959 pba &= E1000_RXPBS_SIZE_MASK_82576; 2960 break; 2961 case e1000_82580: 2962 case e1000_i350: 2963 case e1000_vfadapt_i350: 2964 pba = E1000_READ_REG(hw, E1000_RXPBS); 2965 pba = e1000_rxpbs_adjust_82580(pba); 2966 break; 2967 case e1000_i210: 2968 case e1000_i211: 2969 pba = E1000_PBA_34K; 2970 default: 2971 break; 2972 } 2973 2974 /* Special needs in case of Jumbo frames */ 2975 if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) { 2976 u32 tx_space, min_tx, min_rx; 2977 pba = E1000_READ_REG(hw, E1000_PBA); 2978 tx_space = pba >> 16; 2979 pba &= 0xffff; 2980 min_tx = (adapter->max_frame_size + 2981 sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2; 2982 min_tx = roundup2(min_tx, 1024); 2983 min_tx >>= 10; 2984 min_rx = adapter->max_frame_size; 2985 min_rx = roundup2(min_rx, 1024); 2986 min_rx >>= 10; 2987 if (tx_space < min_tx && 2988 ((min_tx - tx_space) < pba)) { 2989 pba = pba - (min_tx - tx_space); 2990 /* 2991 * if short on rx space, rx wins 2992 * and must trump tx adjustment 2993 */ 2994 if (pba < min_rx) 2995 pba = min_rx; 2996 } 2997 E1000_WRITE_REG(hw, E1000_PBA, pba); 2998 } 2999 3000 INIT_DEBUGOUT1("igb_init: pba=%dK",pba); 3001 3002 /* 3003 * These parameters control the automatic generation (Tx) and 3004 * response (Rx) to Ethernet PAUSE frames. 3005 * - High water mark should allow for at least two frames to be 3006 * received after sending an XOFF. 3007 * - Low water mark works best when it is very near the high water mark. 3008 * This allows the receiver to restart by sending XON when it has 3009 * drained a bit. 3010 */ 3011 hwm = min(((pba << 10) * 9 / 10), 3012 ((pba << 10) - 2 * adapter->max_frame_size)); 3013 3014 if (hw->mac.type < e1000_82576) { 3015 fc->high_water = hwm & 0xFFF8; /* 8-byte granularity */ 3016 fc->low_water = fc->high_water - 8; 3017 } else { 3018 fc->high_water = hwm & 0xFFF0; /* 16-byte granularity */ 3019 fc->low_water = fc->high_water - 16; 3020 } 3021 3022 fc->pause_time = IGB_FC_PAUSE_TIME; 3023 fc->send_xon = TRUE; 3024 if (adapter->fc) 3025 fc->requested_mode = adapter->fc; 3026 else 3027 fc->requested_mode = e1000_fc_default; 3028 3029 /* Issue a global reset */ 3030 e1000_reset_hw(hw); 3031 E1000_WRITE_REG(hw, E1000_WUC, 0); 3032 3033 if (e1000_init_hw(hw) < 0) 3034 device_printf(dev, "Hardware Initialization Failed\n"); 3035 3036 /* Setup DMA Coalescing */ 3037 if ((hw->mac.type > e1000_82580) && 3038 (hw->mac.type != e1000_i211)) { 3039 u32 dmac; 3040 u32 reg = ~E1000_DMACR_DMAC_EN; 3041 3042 if (adapter->dmac == 0) { /* Disabling it */ 3043 E1000_WRITE_REG(hw, E1000_DMACR, reg); 3044 goto reset_out; 3045 } 3046 3047 /* Set starting thresholds */ 3048 E1000_WRITE_REG(hw, E1000_DMCTXTH, 0); 3049 E1000_WRITE_REG(hw, E1000_DMCRTRH, 0); 3050 3051 hwm = 64 * pba - adapter->max_frame_size / 16; 3052 if (hwm < 64 * (pba - 6)) 3053 hwm = 64 * (pba - 6); 3054 reg = E1000_READ_REG(hw, E1000_FCRTC); 3055 reg &= ~E1000_FCRTC_RTH_COAL_MASK; 3056 reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT) 3057 & E1000_FCRTC_RTH_COAL_MASK); 3058 E1000_WRITE_REG(hw, E1000_FCRTC, reg); 3059 3060 3061 dmac = pba - adapter->max_frame_size / 512; 3062 if (dmac < pba - 10) 3063 dmac = pba - 10; 3064 reg = E1000_READ_REG(hw, E1000_DMACR); 3065 reg &= ~E1000_DMACR_DMACTHR_MASK; 3066 reg = ((dmac << E1000_DMACR_DMACTHR_SHIFT) 3067 & E1000_DMACR_DMACTHR_MASK); 3068 /* transition to L0x or L1 if available..*/ 3069 reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK); 3070 /* timer = value in adapter->dmac in 32usec intervals */ 3071 reg |= (adapter->dmac >> 5); 3072 E1000_WRITE_REG(hw, E1000_DMACR, reg); 3073 3074 /* Set the interval before transition */ 3075 reg = E1000_READ_REG(hw, E1000_DMCTLX); 3076 reg |= 0x80000004; 3077 E1000_WRITE_REG(hw, E1000_DMCTLX, reg); 3078 3079 /* free space in tx packet buffer to wake from DMA coal */ 3080 E1000_WRITE_REG(hw, E1000_DMCTXTH, 3081 (20480 - (2 * adapter->max_frame_size)) >> 6); 3082 3083 /* make low power state decision controlled by DMA coal */ 3084 reg = E1000_READ_REG(hw, E1000_PCIEMISC); 3085 reg &= ~E1000_PCIEMISC_LX_DECISION; 3086 E1000_WRITE_REG(hw, E1000_PCIEMISC, reg); 3087 device_printf(dev, "DMA Coalescing enabled\n"); 3088 3089 } else if (hw->mac.type == e1000_82580) { 3090 u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC); 3091 E1000_WRITE_REG(hw, E1000_DMACR, 0); 3092 E1000_WRITE_REG(hw, E1000_PCIEMISC, 3093 reg & ~E1000_PCIEMISC_LX_DECISION); 3094 } 3095 3096reset_out: 3097 E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN); 3098 e1000_get_phy_info(hw); 3099 e1000_check_for_link(hw); 3100 return; 3101} 3102 3103/********************************************************************* 3104 * 3105 * Setup networking device structure and register an interface. 3106 * 3107 **********************************************************************/ 3108static int 3109igb_setup_interface(device_t dev, struct adapter *adapter) 3110{ 3111 struct ifnet *ifp; 3112 3113 INIT_DEBUGOUT("igb_setup_interface: begin"); 3114 3115 ifp = adapter->ifp = if_alloc(IFT_ETHER); 3116 if (ifp == NULL) { 3117 device_printf(dev, "can not allocate ifnet structure\n"); 3118 return (-1); 3119 } 3120 if_initname(ifp, device_get_name(dev), device_get_unit(dev)); 3121 ifp->if_init = igb_init; 3122 ifp->if_softc = adapter; 3123 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 3124 ifp->if_ioctl = igb_ioctl; 3125#ifndef IGB_LEGACY_TX 3126 ifp->if_transmit = igb_mq_start; 3127 ifp->if_qflush = igb_qflush; 3128#else 3129 ifp->if_start = igb_start; 3130 IFQ_SET_MAXLEN(&ifp->if_snd, adapter->num_tx_desc - 1); 3131 ifp->if_snd.ifq_drv_maxlen = adapter->num_tx_desc - 1; 3132 IFQ_SET_READY(&ifp->if_snd); 3133#endif 3134 3135 ether_ifattach(ifp, adapter->hw.mac.addr); 3136 3137 ifp->if_capabilities = ifp->if_capenable = 0; 3138 3139 ifp->if_capabilities = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM; 3140 ifp->if_capabilities |= IFCAP_TSO4; 3141 ifp->if_capabilities |= IFCAP_JUMBO_MTU; 3142 ifp->if_capenable = ifp->if_capabilities; 3143 3144 /* Don't enable LRO by default */ 3145 ifp->if_capabilities |= IFCAP_LRO; 3146 3147#ifdef DEVICE_POLLING 3148 ifp->if_capabilities |= IFCAP_POLLING; 3149#endif 3150 3151 /* 3152 * Tell the upper layer(s) we 3153 * support full VLAN capability. 3154 */ 3155 ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header); 3156 ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING 3157 | IFCAP_VLAN_HWTSO 3158 | IFCAP_VLAN_MTU; 3159 ifp->if_capenable |= IFCAP_VLAN_HWTAGGING 3160 | IFCAP_VLAN_HWTSO 3161 | IFCAP_VLAN_MTU; 3162 3163 /* 3164 ** Don't turn this on by default, if vlans are 3165 ** created on another pseudo device (eg. lagg) 3166 ** then vlan events are not passed thru, breaking 3167 ** operation, but with HW FILTER off it works. If 3168 ** using vlans directly on the igb driver you can 3169 ** enable this and get full hardware tag filtering. 3170 */ 3171 ifp->if_capabilities |= IFCAP_VLAN_HWFILTER; 3172 3173 /* 3174 * Specify the media types supported by this adapter and register 3175 * callbacks to update media and link information 3176 */ 3177 ifmedia_init(&adapter->media, IFM_IMASK, 3178 igb_media_change, igb_media_status); 3179 if ((adapter->hw.phy.media_type == e1000_media_type_fiber) || 3180 (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) { 3181 ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_SX | IFM_FDX, 3182 0, NULL); 3183 ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_SX, 0, NULL); 3184 } else { 3185 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T, 0, NULL); 3186 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 3187 0, NULL); 3188 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX, 3189 0, NULL); 3190 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 3191 0, NULL); 3192 if (adapter->hw.phy.type != e1000_phy_ife) { 3193 ifmedia_add(&adapter->media, 3194 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL); 3195 ifmedia_add(&adapter->media, 3196 IFM_ETHER | IFM_1000_T, 0, NULL); 3197 } 3198 } 3199 ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL); 3200 ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO); 3201 return (0); 3202} 3203 3204 3205/* 3206 * Manage DMA'able memory. 3207 */ 3208static void 3209igb_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error) 3210{ 3211 if (error) 3212 return; 3213 *(bus_addr_t *) arg = segs[0].ds_addr; 3214} 3215 3216static int 3217igb_dma_malloc(struct adapter *adapter, bus_size_t size, 3218 struct igb_dma_alloc *dma, int mapflags) 3219{ 3220 int error; 3221 3222 error = bus_dma_tag_create(bus_get_dma_tag(adapter->dev), /* parent */ 3223 IGB_DBA_ALIGN, 0, /* alignment, bounds */ 3224 BUS_SPACE_MAXADDR, /* lowaddr */ 3225 BUS_SPACE_MAXADDR, /* highaddr */ 3226 NULL, NULL, /* filter, filterarg */ 3227 size, /* maxsize */ 3228 1, /* nsegments */ 3229 size, /* maxsegsize */ 3230 0, /* flags */ 3231 NULL, /* lockfunc */ 3232 NULL, /* lockarg */ 3233 &dma->dma_tag); 3234 if (error) { 3235 device_printf(adapter->dev, 3236 "%s: bus_dma_tag_create failed: %d\n", 3237 __func__, error); 3238 goto fail_0; 3239 } 3240 3241 error = bus_dmamem_alloc(dma->dma_tag, (void**) &dma->dma_vaddr, 3242 BUS_DMA_NOWAIT | BUS_DMA_COHERENT, &dma->dma_map); 3243 if (error) { 3244 device_printf(adapter->dev, 3245 "%s: bus_dmamem_alloc(%ju) failed: %d\n", 3246 __func__, (uintmax_t)size, error); 3247 goto fail_2; 3248 } 3249 3250 dma->dma_paddr = 0; 3251 error = bus_dmamap_load(dma->dma_tag, dma->dma_map, dma->dma_vaddr, 3252 size, igb_dmamap_cb, &dma->dma_paddr, mapflags | BUS_DMA_NOWAIT); 3253 if (error || dma->dma_paddr == 0) { 3254 device_printf(adapter->dev, 3255 "%s: bus_dmamap_load failed: %d\n", 3256 __func__, error); 3257 goto fail_3; 3258 } 3259 3260 return (0); 3261 3262fail_3: 3263 bus_dmamap_unload(dma->dma_tag, dma->dma_map); 3264fail_2: 3265 bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map); 3266 bus_dma_tag_destroy(dma->dma_tag); 3267fail_0: 3268 dma->dma_map = NULL; 3269 dma->dma_tag = NULL; 3270 3271 return (error); 3272} 3273 3274static void 3275igb_dma_free(struct adapter *adapter, struct igb_dma_alloc *dma) 3276{ 3277 if (dma->dma_tag == NULL) 3278 return; 3279 if (dma->dma_map != NULL) { 3280 bus_dmamap_sync(dma->dma_tag, dma->dma_map, 3281 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 3282 bus_dmamap_unload(dma->dma_tag, dma->dma_map); 3283 bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map); 3284 dma->dma_map = NULL; 3285 } 3286 bus_dma_tag_destroy(dma->dma_tag); 3287 dma->dma_tag = NULL; 3288} 3289 3290 3291/********************************************************************* 3292 * 3293 * Allocate memory for the transmit and receive rings, and then 3294 * the descriptors associated with each, called only once at attach. 3295 * 3296 **********************************************************************/ 3297static int 3298igb_allocate_queues(struct adapter *adapter) 3299{ 3300 device_t dev = adapter->dev; 3301 struct igb_queue *que = NULL; 3302 struct tx_ring *txr = NULL; 3303 struct rx_ring *rxr = NULL; 3304 int rsize, tsize, error = E1000_SUCCESS; 3305 int txconf = 0, rxconf = 0; 3306 3307 /* First allocate the top level queue structs */ 3308 if (!(adapter->queues = 3309 (struct igb_queue *) malloc(sizeof(struct igb_queue) * 3310 adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { 3311 device_printf(dev, "Unable to allocate queue memory\n"); 3312 error = ENOMEM; 3313 goto fail; 3314 } 3315 3316 /* Next allocate the TX ring struct memory */ 3317 if (!(adapter->tx_rings = 3318 (struct tx_ring *) malloc(sizeof(struct tx_ring) * 3319 adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { 3320 device_printf(dev, "Unable to allocate TX ring memory\n"); 3321 error = ENOMEM; 3322 goto tx_fail; 3323 } 3324 3325 /* Now allocate the RX */ 3326 if (!(adapter->rx_rings = 3327 (struct rx_ring *) malloc(sizeof(struct rx_ring) * 3328 adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { 3329 device_printf(dev, "Unable to allocate RX ring memory\n"); 3330 error = ENOMEM; 3331 goto rx_fail; 3332 } 3333 3334 tsize = roundup2(adapter->num_tx_desc * 3335 sizeof(union e1000_adv_tx_desc), IGB_DBA_ALIGN); 3336 /* 3337 * Now set up the TX queues, txconf is needed to handle the 3338 * possibility that things fail midcourse and we need to 3339 * undo memory gracefully 3340 */ 3341 for (int i = 0; i < adapter->num_queues; i++, txconf++) { 3342 /* Set up some basics */ 3343 txr = &adapter->tx_rings[i]; 3344 txr->adapter = adapter; 3345 txr->me = i; 3346 3347 /* Initialize the TX lock */ 3348 snprintf(txr->mtx_name, sizeof(txr->mtx_name), "%s:tx(%d)", 3349 device_get_nameunit(dev), txr->me); 3350 mtx_init(&txr->tx_mtx, txr->mtx_name, NULL, MTX_DEF); 3351 3352 if (igb_dma_malloc(adapter, tsize, 3353 &txr->txdma, BUS_DMA_NOWAIT)) { 3354 device_printf(dev, 3355 "Unable to allocate TX Descriptor memory\n"); 3356 error = ENOMEM; 3357 goto err_tx_desc; 3358 } 3359 txr->tx_base = (struct e1000_tx_desc *)txr->txdma.dma_vaddr; 3360 bzero((void *)txr->tx_base, tsize); 3361 3362 /* Now allocate transmit buffers for the ring */ 3363 if (igb_allocate_transmit_buffers(txr)) { 3364 device_printf(dev, 3365 "Critical Failure setting up transmit buffers\n"); 3366 error = ENOMEM; 3367 goto err_tx_desc; 3368 } 3369#ifndef IGB_LEGACY_TX 3370 /* Allocate a buf ring */ 3371 txr->br = buf_ring_alloc(igb_buf_ring_size, M_DEVBUF, 3372 M_WAITOK, &txr->tx_mtx); 3373#endif 3374 } 3375 3376 /* 3377 * Next the RX queues... 3378 */ 3379 rsize = roundup2(adapter->num_rx_desc * 3380 sizeof(union e1000_adv_rx_desc), IGB_DBA_ALIGN); 3381 for (int i = 0; i < adapter->num_queues; i++, rxconf++) { 3382 rxr = &adapter->rx_rings[i]; 3383 rxr->adapter = adapter; 3384 rxr->me = i; 3385 3386 /* Initialize the RX lock */ 3387 snprintf(rxr->mtx_name, sizeof(rxr->mtx_name), "%s:rx(%d)", 3388 device_get_nameunit(dev), txr->me); 3389 mtx_init(&rxr->rx_mtx, rxr->mtx_name, NULL, MTX_DEF); 3390 3391 if (igb_dma_malloc(adapter, rsize, 3392 &rxr->rxdma, BUS_DMA_NOWAIT)) { 3393 device_printf(dev, 3394 "Unable to allocate RxDescriptor memory\n"); 3395 error = ENOMEM; 3396 goto err_rx_desc; 3397 } 3398 rxr->rx_base = (union e1000_adv_rx_desc *)rxr->rxdma.dma_vaddr; 3399 bzero((void *)rxr->rx_base, rsize); 3400 3401 /* Allocate receive buffers for the ring*/ 3402 if (igb_allocate_receive_buffers(rxr)) { 3403 device_printf(dev, 3404 "Critical Failure setting up receive buffers\n"); 3405 error = ENOMEM; 3406 goto err_rx_desc; 3407 } 3408 } 3409 3410 /* 3411 ** Finally set up the queue holding structs 3412 */ 3413 for (int i = 0; i < adapter->num_queues; i++) { 3414 que = &adapter->queues[i]; 3415 que->adapter = adapter; 3416 que->txr = &adapter->tx_rings[i]; 3417 que->rxr = &adapter->rx_rings[i]; 3418 } 3419 3420 return (0); 3421 3422err_rx_desc: 3423 for (rxr = adapter->rx_rings; rxconf > 0; rxr++, rxconf--) 3424 igb_dma_free(adapter, &rxr->rxdma); 3425err_tx_desc: 3426 for (txr = adapter->tx_rings; txconf > 0; txr++, txconf--) 3427 igb_dma_free(adapter, &txr->txdma); 3428 free(adapter->rx_rings, M_DEVBUF); 3429rx_fail: 3430#ifndef IGB_LEGACY_TX 3431 buf_ring_free(txr->br, M_DEVBUF); 3432#endif 3433 free(adapter->tx_rings, M_DEVBUF); 3434tx_fail: 3435 free(adapter->queues, M_DEVBUF); 3436fail: 3437 return (error); 3438} 3439 3440/********************************************************************* 3441 * 3442 * Allocate memory for tx_buffer structures. The tx_buffer stores all 3443 * the information needed to transmit a packet on the wire. This is 3444 * called only once at attach, setup is done every reset. 3445 * 3446 **********************************************************************/ 3447static int 3448igb_allocate_transmit_buffers(struct tx_ring *txr) 3449{ 3450 struct adapter *adapter = txr->adapter; 3451 device_t dev = adapter->dev; 3452 struct igb_tx_buffer *txbuf; 3453 int error, i; 3454 3455 /* 3456 * Setup DMA descriptor areas. 3457 */ 3458 if ((error = bus_dma_tag_create(bus_get_dma_tag(dev), 3459 1, 0, /* alignment, bounds */ 3460 BUS_SPACE_MAXADDR, /* lowaddr */ 3461 BUS_SPACE_MAXADDR, /* highaddr */ 3462 NULL, NULL, /* filter, filterarg */ 3463 IGB_TSO_SIZE, /* maxsize */ 3464 IGB_MAX_SCATTER, /* nsegments */ 3465 PAGE_SIZE, /* maxsegsize */ 3466 0, /* flags */ 3467 NULL, /* lockfunc */ 3468 NULL, /* lockfuncarg */ 3469 &txr->txtag))) { 3470 device_printf(dev,"Unable to allocate TX DMA tag\n"); 3471 goto fail; 3472 } 3473 3474 if (!(txr->tx_buffers = 3475 (struct igb_tx_buffer *) malloc(sizeof(struct igb_tx_buffer) * 3476 adapter->num_tx_desc, M_DEVBUF, M_NOWAIT | M_ZERO))) { 3477 device_printf(dev, "Unable to allocate tx_buffer memory\n"); 3478 error = ENOMEM; 3479 goto fail; 3480 } 3481 3482 /* Create the descriptor buffer dma maps */ 3483 txbuf = txr->tx_buffers; 3484 for (i = 0; i < adapter->num_tx_desc; i++, txbuf++) { 3485 error = bus_dmamap_create(txr->txtag, 0, &txbuf->map); 3486 if (error != 0) { 3487 device_printf(dev, "Unable to create TX DMA map\n"); 3488 goto fail; 3489 } 3490 } 3491 3492 return 0; 3493fail: 3494 /* We free all, it handles case where we are in the middle */ 3495 igb_free_transmit_structures(adapter); 3496 return (error); 3497} 3498 3499/********************************************************************* 3500 * 3501 * Initialize a transmit ring. 3502 * 3503 **********************************************************************/ 3504static void 3505igb_setup_transmit_ring(struct tx_ring *txr) 3506{ 3507 struct adapter *adapter = txr->adapter; 3508 struct igb_tx_buffer *txbuf; 3509 int i; 3510#ifdef DEV_NETMAP 3511 struct netmap_adapter *na = NA(adapter->ifp); 3512 struct netmap_slot *slot; 3513#endif /* DEV_NETMAP */ 3514 3515 /* Clear the old descriptor contents */ 3516 IGB_TX_LOCK(txr); 3517#ifdef DEV_NETMAP 3518 slot = netmap_reset(na, NR_TX, txr->me, 0); 3519#endif /* DEV_NETMAP */ 3520 bzero((void *)txr->tx_base, 3521 (sizeof(union e1000_adv_tx_desc)) * adapter->num_tx_desc); 3522 /* Reset indices */ 3523 txr->next_avail_desc = 0; 3524 txr->next_to_clean = 0; 3525 3526 /* Free any existing tx buffers. */ 3527 txbuf = txr->tx_buffers; 3528 for (i = 0; i < adapter->num_tx_desc; i++, txbuf++) { 3529 if (txbuf->m_head != NULL) { 3530 bus_dmamap_sync(txr->txtag, txbuf->map, 3531 BUS_DMASYNC_POSTWRITE); 3532 bus_dmamap_unload(txr->txtag, txbuf->map); 3533 m_freem(txbuf->m_head); 3534 txbuf->m_head = NULL; 3535 } 3536#ifdef DEV_NETMAP 3537 if (slot) { 3538 int si = netmap_idx_n2k(&na->tx_rings[txr->me], i); 3539 /* no need to set the address */ 3540 netmap_load_map(txr->txtag, txbuf->map, NMB(slot + si)); 3541 } 3542#endif /* DEV_NETMAP */ 3543 /* clear the watch index */ 3544 txbuf->next_eop = -1; 3545 } 3546 3547 /* Set number of descriptors available */ 3548 txr->tx_avail = adapter->num_tx_desc; 3549 3550 bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, 3551 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 3552 IGB_TX_UNLOCK(txr); 3553} 3554 3555/********************************************************************* 3556 * 3557 * Initialize all transmit rings. 3558 * 3559 **********************************************************************/ 3560static void 3561igb_setup_transmit_structures(struct adapter *adapter) 3562{ 3563 struct tx_ring *txr = adapter->tx_rings; 3564 3565 for (int i = 0; i < adapter->num_queues; i++, txr++) 3566 igb_setup_transmit_ring(txr); 3567 3568 return; 3569} 3570 3571/********************************************************************* 3572 * 3573 * Enable transmit unit. 3574 * 3575 **********************************************************************/ 3576static void 3577igb_initialize_transmit_units(struct adapter *adapter) 3578{ 3579 struct tx_ring *txr = adapter->tx_rings; 3580 struct e1000_hw *hw = &adapter->hw; 3581 u32 tctl, txdctl; 3582 3583 INIT_DEBUGOUT("igb_initialize_transmit_units: begin"); 3584 tctl = txdctl = 0; 3585 3586 /* Setup the Tx Descriptor Rings */ 3587 for (int i = 0; i < adapter->num_queues; i++, txr++) { 3588 u64 bus_addr = txr->txdma.dma_paddr; 3589 3590 E1000_WRITE_REG(hw, E1000_TDLEN(i), 3591 adapter->num_tx_desc * sizeof(struct e1000_tx_desc)); 3592 E1000_WRITE_REG(hw, E1000_TDBAH(i), 3593 (uint32_t)(bus_addr >> 32)); 3594 E1000_WRITE_REG(hw, E1000_TDBAL(i), 3595 (uint32_t)bus_addr); 3596 3597 /* Setup the HW Tx Head and Tail descriptor pointers */ 3598 E1000_WRITE_REG(hw, E1000_TDT(i), 0); 3599 E1000_WRITE_REG(hw, E1000_TDH(i), 0); 3600 3601 HW_DEBUGOUT2("Base = %x, Length = %x\n", 3602 E1000_READ_REG(hw, E1000_TDBAL(i)), 3603 E1000_READ_REG(hw, E1000_TDLEN(i))); 3604 3605 txr->queue_status = IGB_QUEUE_IDLE; 3606 3607 txdctl |= IGB_TX_PTHRESH; 3608 txdctl |= IGB_TX_HTHRESH << 8; 3609 txdctl |= IGB_TX_WTHRESH << 16; 3610 txdctl |= E1000_TXDCTL_QUEUE_ENABLE; 3611 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl); 3612 } 3613 3614 if (adapter->vf_ifp) 3615 return; 3616 3617 e1000_config_collision_dist(hw); 3618 3619 /* Program the Transmit Control Register */ 3620 tctl = E1000_READ_REG(hw, E1000_TCTL); 3621 tctl &= ~E1000_TCTL_CT; 3622 tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN | 3623 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT)); 3624 3625 /* This write will effectively turn on the transmit unit. */ 3626 E1000_WRITE_REG(hw, E1000_TCTL, tctl); 3627} 3628 3629/********************************************************************* 3630 * 3631 * Free all transmit rings. 3632 * 3633 **********************************************************************/ 3634static void 3635igb_free_transmit_structures(struct adapter *adapter) 3636{ 3637 struct tx_ring *txr = adapter->tx_rings; 3638 3639 for (int i = 0; i < adapter->num_queues; i++, txr++) { 3640 IGB_TX_LOCK(txr); 3641 igb_free_transmit_buffers(txr); 3642 igb_dma_free(adapter, &txr->txdma); 3643 IGB_TX_UNLOCK(txr); 3644 IGB_TX_LOCK_DESTROY(txr); 3645 } 3646 free(adapter->tx_rings, M_DEVBUF); 3647} 3648 3649/********************************************************************* 3650 * 3651 * Free transmit ring related data structures. 3652 * 3653 **********************************************************************/ 3654static void 3655igb_free_transmit_buffers(struct tx_ring *txr) 3656{ 3657 struct adapter *adapter = txr->adapter; 3658 struct igb_tx_buffer *tx_buffer; 3659 int i; 3660 3661 INIT_DEBUGOUT("free_transmit_ring: begin"); 3662 3663 if (txr->tx_buffers == NULL) 3664 return; 3665 3666 tx_buffer = txr->tx_buffers; 3667 for (i = 0; i < adapter->num_tx_desc; i++, tx_buffer++) { 3668 if (tx_buffer->m_head != NULL) { 3669 bus_dmamap_sync(txr->txtag, tx_buffer->map, 3670 BUS_DMASYNC_POSTWRITE); 3671 bus_dmamap_unload(txr->txtag, 3672 tx_buffer->map); 3673 m_freem(tx_buffer->m_head); 3674 tx_buffer->m_head = NULL; 3675 if (tx_buffer->map != NULL) { 3676 bus_dmamap_destroy(txr->txtag, 3677 tx_buffer->map); 3678 tx_buffer->map = NULL; 3679 } 3680 } else if (tx_buffer->map != NULL) { 3681 bus_dmamap_unload(txr->txtag, 3682 tx_buffer->map); 3683 bus_dmamap_destroy(txr->txtag, 3684 tx_buffer->map); 3685 tx_buffer->map = NULL; 3686 } 3687 } 3688#ifndef IGB_LEGACY_TX 3689 if (txr->br != NULL) 3690 buf_ring_free(txr->br, M_DEVBUF); 3691#endif 3692 if (txr->tx_buffers != NULL) { 3693 free(txr->tx_buffers, M_DEVBUF); 3694 txr->tx_buffers = NULL; 3695 } 3696 if (txr->txtag != NULL) { 3697 bus_dma_tag_destroy(txr->txtag); 3698 txr->txtag = NULL; 3699 } 3700 return; 3701} 3702 3703/********************************************************************** 3704 * 3705 * Setup work for hardware segmentation offload (TSO) 3706 * 3707 **********************************************************************/ 3708static bool 3709igb_tso_setup(struct tx_ring *txr, struct mbuf *mp, int ehdrlen, 3710 struct ip *ip, struct tcphdr *th) 3711{ 3712 struct adapter *adapter = txr->adapter; 3713 struct e1000_adv_tx_context_desc *TXD; 3714 struct igb_tx_buffer *tx_buffer; 3715 u32 vlan_macip_lens = 0, type_tucmd_mlhl = 0; 3716 u32 mss_l4len_idx = 0; 3717 u16 vtag = 0; 3718 int ctxd, ip_hlen, tcp_hlen; 3719 3720 ctxd = txr->next_avail_desc; 3721 tx_buffer = &txr->tx_buffers[ctxd]; 3722 TXD = (struct e1000_adv_tx_context_desc *) &txr->tx_base[ctxd]; 3723 3724 ip->ip_sum = 0; 3725 ip_hlen = ip->ip_hl << 2; 3726 tcp_hlen = th->th_off << 2; 3727 3728 /* VLAN MACLEN IPLEN */ 3729 if (mp->m_flags & M_VLANTAG) { 3730 vtag = htole16(mp->m_pkthdr.ether_vtag); 3731 vlan_macip_lens |= (vtag << E1000_ADVTXD_VLAN_SHIFT); 3732 } 3733 3734 vlan_macip_lens |= (ehdrlen << E1000_ADVTXD_MACLEN_SHIFT); 3735 vlan_macip_lens |= ip_hlen; 3736 TXD->vlan_macip_lens |= htole32(vlan_macip_lens); 3737 3738 /* ADV DTYPE TUCMD */ 3739 type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT; 3740 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP; 3741 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4; 3742 TXD->type_tucmd_mlhl |= htole32(type_tucmd_mlhl); 3743 3744 /* MSS L4LEN IDX */ 3745 mss_l4len_idx |= (mp->m_pkthdr.tso_segsz << E1000_ADVTXD_MSS_SHIFT); 3746 mss_l4len_idx |= (tcp_hlen << E1000_ADVTXD_L4LEN_SHIFT); 3747 /* 82575 needs the queue index added */ 3748 if (adapter->hw.mac.type == e1000_82575) 3749 mss_l4len_idx |= txr->me << 4; 3750 TXD->mss_l4len_idx = htole32(mss_l4len_idx); 3751 3752 TXD->seqnum_seed = htole32(0); 3753 tx_buffer->m_head = NULL; 3754 tx_buffer->next_eop = -1; 3755 3756 if (++ctxd == adapter->num_tx_desc) 3757 ctxd = 0; 3758 3759 txr->tx_avail--; 3760 txr->next_avail_desc = ctxd; 3761 return TRUE; 3762} 3763 3764 3765/********************************************************************* 3766 * 3767 * Context Descriptor setup for VLAN or CSUM 3768 * 3769 **********************************************************************/ 3770 3771static bool 3772igb_tx_ctx_setup(struct tx_ring *txr, struct mbuf *mp) 3773{ 3774 struct adapter *adapter = txr->adapter; 3775 struct e1000_adv_tx_context_desc *TXD; 3776 struct igb_tx_buffer *tx_buffer; 3777 u32 vlan_macip_lens, type_tucmd_mlhl, mss_l4len_idx; 3778 struct ether_vlan_header *eh; 3779 struct ip *ip = NULL; 3780 struct ip6_hdr *ip6; 3781 int ehdrlen, ctxd, ip_hlen = 0; 3782 u16 etype, vtag = 0; 3783 u8 ipproto = 0; 3784 bool offload = TRUE; 3785 3786 if ((mp->m_pkthdr.csum_flags & CSUM_OFFLOAD) == 0) 3787 offload = FALSE; 3788 3789 vlan_macip_lens = type_tucmd_mlhl = mss_l4len_idx = 0; 3790 ctxd = txr->next_avail_desc; 3791 tx_buffer = &txr->tx_buffers[ctxd]; 3792 TXD = (struct e1000_adv_tx_context_desc *) &txr->tx_base[ctxd]; 3793 3794 /* 3795 ** In advanced descriptors the vlan tag must 3796 ** be placed into the context descriptor, thus 3797 ** we need to be here just for that setup. 3798 */ 3799 if (mp->m_flags & M_VLANTAG) { 3800 vtag = htole16(mp->m_pkthdr.ether_vtag); 3801 vlan_macip_lens |= (vtag << E1000_ADVTXD_VLAN_SHIFT); 3802 } else if (offload == FALSE) 3803 return FALSE; 3804 3805 /* 3806 * Determine where frame payload starts. 3807 * Jump over vlan headers if already present, 3808 * helpful for QinQ too. 3809 */ 3810 eh = mtod(mp, struct ether_vlan_header *); 3811 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { 3812 etype = ntohs(eh->evl_proto); 3813 ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN; 3814 } else { 3815 etype = ntohs(eh->evl_encap_proto); 3816 ehdrlen = ETHER_HDR_LEN; 3817 } 3818 3819 /* Set the ether header length */ 3820 vlan_macip_lens |= ehdrlen << E1000_ADVTXD_MACLEN_SHIFT; 3821 3822 switch (etype) { 3823 case ETHERTYPE_IP: 3824 ip = (struct ip *)(mp->m_data + ehdrlen); 3825 ip_hlen = ip->ip_hl << 2; 3826 if (mp->m_len < ehdrlen + ip_hlen) { 3827 offload = FALSE; 3828 break; 3829 } 3830 ipproto = ip->ip_p; 3831 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4; 3832 break; 3833 case ETHERTYPE_IPV6: 3834 ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen); 3835 ip_hlen = sizeof(struct ip6_hdr); 3836 ipproto = ip6->ip6_nxt; 3837 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV6; 3838 break; 3839 default: 3840 offload = FALSE; 3841 break; 3842 } 3843 3844 vlan_macip_lens |= ip_hlen; 3845 type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT; 3846 3847 switch (ipproto) { 3848 case IPPROTO_TCP: 3849 if (mp->m_pkthdr.csum_flags & CSUM_TCP) 3850 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP; 3851 break; 3852 case IPPROTO_UDP: 3853 if (mp->m_pkthdr.csum_flags & CSUM_UDP) 3854 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_UDP; 3855 break; 3856#if __FreeBSD_version >= 800000 3857 case IPPROTO_SCTP: 3858 if (mp->m_pkthdr.csum_flags & CSUM_SCTP) 3859 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_SCTP; 3860 break; 3861#endif 3862 default: 3863 offload = FALSE; 3864 break; 3865 } 3866 3867 /* 82575 needs the queue index added */ 3868 if (adapter->hw.mac.type == e1000_82575) 3869 mss_l4len_idx = txr->me << 4; 3870 3871 /* Now copy bits into descriptor */ 3872 TXD->vlan_macip_lens |= htole32(vlan_macip_lens); 3873 TXD->type_tucmd_mlhl |= htole32(type_tucmd_mlhl); 3874 TXD->seqnum_seed = htole32(0); 3875 TXD->mss_l4len_idx = htole32(mss_l4len_idx); 3876 3877 tx_buffer->m_head = NULL; 3878 tx_buffer->next_eop = -1; 3879 3880 /* We've consumed the first desc, adjust counters */ 3881 if (++ctxd == adapter->num_tx_desc) 3882 ctxd = 0; 3883 txr->next_avail_desc = ctxd; 3884 --txr->tx_avail; 3885 3886 return (offload); 3887} 3888 3889 3890/********************************************************************** 3891 * 3892 * Examine each tx_buffer in the used queue. If the hardware is done 3893 * processing the packet then free associated resources. The 3894 * tx_buffer is put back on the free queue. 3895 * 3896 * TRUE return means there's work in the ring to clean, FALSE its empty. 3897 **********************************************************************/ 3898static bool 3899igb_txeof(struct tx_ring *txr) 3900{ 3901 struct adapter *adapter = txr->adapter; 3902 int first, last, done, processed; 3903 struct igb_tx_buffer *tx_buffer; 3904 struct e1000_tx_desc *tx_desc, *eop_desc; 3905 struct ifnet *ifp = adapter->ifp; 3906 3907 IGB_TX_LOCK_ASSERT(txr); 3908 3909#ifdef DEV_NETMAP 3910 if (netmap_tx_irq(ifp, txr->me | 3911 (NETMAP_LOCKED_ENTER|NETMAP_LOCKED_EXIT))) 3912 return (FALSE); 3913#endif /* DEV_NETMAP */ 3914 if (txr->tx_avail == adapter->num_tx_desc) { 3915 txr->queue_status = IGB_QUEUE_IDLE; 3916 return FALSE; 3917 } 3918 3919 processed = 0; 3920 first = txr->next_to_clean; 3921 tx_desc = &txr->tx_base[first]; 3922 tx_buffer = &txr->tx_buffers[first]; 3923 last = tx_buffer->next_eop; 3924 eop_desc = &txr->tx_base[last]; 3925 3926 /* 3927 * What this does is get the index of the 3928 * first descriptor AFTER the EOP of the 3929 * first packet, that way we can do the 3930 * simple comparison on the inner while loop. 3931 */ 3932 if (++last == adapter->num_tx_desc) 3933 last = 0; 3934 done = last; 3935 3936 bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, 3937 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 3938 3939 while (eop_desc->upper.fields.status & E1000_TXD_STAT_DD) { 3940 /* We clean the range of the packet */ 3941 while (first != done) { 3942 tx_desc->upper.data = 0; 3943 tx_desc->lower.data = 0; 3944 tx_desc->buffer_addr = 0; 3945 ++txr->tx_avail; 3946 ++processed; 3947 3948 if (tx_buffer->m_head) { 3949 txr->bytes += 3950 tx_buffer->m_head->m_pkthdr.len; 3951 bus_dmamap_sync(txr->txtag, 3952 tx_buffer->map, 3953 BUS_DMASYNC_POSTWRITE); 3954 bus_dmamap_unload(txr->txtag, 3955 tx_buffer->map); 3956 3957 m_freem(tx_buffer->m_head); 3958 tx_buffer->m_head = NULL; 3959 } 3960 tx_buffer->next_eop = -1; 3961 txr->watchdog_time = ticks; 3962 3963 if (++first == adapter->num_tx_desc) 3964 first = 0; 3965 3966 tx_buffer = &txr->tx_buffers[first]; 3967 tx_desc = &txr->tx_base[first]; 3968 } 3969 ++txr->packets; 3970 ++ifp->if_opackets; 3971 /* See if we can continue to the next packet */ 3972 last = tx_buffer->next_eop; 3973 if (last != -1) { 3974 eop_desc = &txr->tx_base[last]; 3975 /* Get new done point */ 3976 if (++last == adapter->num_tx_desc) last = 0; 3977 done = last; 3978 } else 3979 break; 3980 } 3981 bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, 3982 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 3983 3984 txr->next_to_clean = first; 3985 3986 /* 3987 ** Watchdog calculation, we know there's 3988 ** work outstanding or the first return 3989 ** would have been taken, so none processed 3990 ** for too long indicates a hang. 3991 */ 3992 if ((!processed) && ((ticks - txr->watchdog_time) > IGB_WATCHDOG)) 3993 txr->queue_status |= IGB_QUEUE_HUNG; 3994 /* 3995 * If we have a minimum free, 3996 * clear depleted state bit 3997 */ 3998 if (txr->tx_avail >= IGB_QUEUE_THRESHOLD) 3999 txr->queue_status &= ~IGB_QUEUE_DEPLETED; 4000 4001 /* All clean, turn off the watchdog */ 4002 if (txr->tx_avail == adapter->num_tx_desc) { 4003 txr->queue_status = IGB_QUEUE_IDLE; 4004 return (FALSE); 4005 } 4006 4007 return (TRUE); 4008} 4009 4010/********************************************************************* 4011 * 4012 * Refresh mbuf buffers for RX descriptor rings 4013 * - now keeps its own state so discards due to resource 4014 * exhaustion are unnecessary, if an mbuf cannot be obtained 4015 * it just returns, keeping its placeholder, thus it can simply 4016 * be recalled to try again. 4017 * 4018 **********************************************************************/ 4019static void 4020igb_refresh_mbufs(struct rx_ring *rxr, int limit) 4021{ 4022 struct adapter *adapter = rxr->adapter; 4023 bus_dma_segment_t hseg[1]; 4024 bus_dma_segment_t pseg[1]; 4025 struct igb_rx_buf *rxbuf; 4026 struct mbuf *mh, *mp; 4027 int i, j, nsegs, error; 4028 bool refreshed = FALSE; 4029 4030 i = j = rxr->next_to_refresh; 4031 /* 4032 ** Get one descriptor beyond 4033 ** our work mark to control 4034 ** the loop. 4035 */ 4036 if (++j == adapter->num_rx_desc) 4037 j = 0; 4038 4039 while (j != limit) { 4040 rxbuf = &rxr->rx_buffers[i]; 4041 /* No hdr mbuf used with header split off */ 4042 if (rxr->hdr_split == FALSE) 4043 goto no_split; 4044 if (rxbuf->m_head == NULL) { 4045 mh = m_gethdr(M_NOWAIT, MT_DATA); 4046 if (mh == NULL) 4047 goto update; 4048 } else 4049 mh = rxbuf->m_head; 4050 4051 mh->m_pkthdr.len = mh->m_len = MHLEN; 4052 mh->m_len = MHLEN; 4053 mh->m_flags |= M_PKTHDR; 4054 /* Get the memory mapping */ 4055 error = bus_dmamap_load_mbuf_sg(rxr->htag, 4056 rxbuf->hmap, mh, hseg, &nsegs, BUS_DMA_NOWAIT); 4057 if (error != 0) { 4058 printf("Refresh mbufs: hdr dmamap load" 4059 " failure - %d\n", error); 4060 m_free(mh); 4061 rxbuf->m_head = NULL; 4062 goto update; 4063 } 4064 rxbuf->m_head = mh; 4065 bus_dmamap_sync(rxr->htag, rxbuf->hmap, 4066 BUS_DMASYNC_PREREAD); 4067 rxr->rx_base[i].read.hdr_addr = 4068 htole64(hseg[0].ds_addr); 4069no_split: 4070 if (rxbuf->m_pack == NULL) { 4071 mp = m_getjcl(M_NOWAIT, MT_DATA, 4072 M_PKTHDR, adapter->rx_mbuf_sz); 4073 if (mp == NULL) 4074 goto update; 4075 } else 4076 mp = rxbuf->m_pack; 4077 4078 mp->m_pkthdr.len = mp->m_len = adapter->rx_mbuf_sz; 4079 /* Get the memory mapping */ 4080 error = bus_dmamap_load_mbuf_sg(rxr->ptag, 4081 rxbuf->pmap, mp, pseg, &nsegs, BUS_DMA_NOWAIT); 4082 if (error != 0) { 4083 printf("Refresh mbufs: payload dmamap load" 4084 " failure - %d\n", error); 4085 m_free(mp); 4086 rxbuf->m_pack = NULL; 4087 goto update; 4088 } 4089 rxbuf->m_pack = mp; 4090 bus_dmamap_sync(rxr->ptag, rxbuf->pmap, 4091 BUS_DMASYNC_PREREAD); 4092 rxr->rx_base[i].read.pkt_addr = 4093 htole64(pseg[0].ds_addr); 4094 refreshed = TRUE; /* I feel wefreshed :) */ 4095 4096 i = j; /* our next is precalculated */ 4097 rxr->next_to_refresh = i; 4098 if (++j == adapter->num_rx_desc) 4099 j = 0; 4100 } 4101update: 4102 if (refreshed) /* update tail */ 4103 E1000_WRITE_REG(&adapter->hw, 4104 E1000_RDT(rxr->me), rxr->next_to_refresh); 4105 return; 4106} 4107 4108 4109/********************************************************************* 4110 * 4111 * Allocate memory for rx_buffer structures. Since we use one 4112 * rx_buffer per received packet, the maximum number of rx_buffer's 4113 * that we'll need is equal to the number of receive descriptors 4114 * that we've allocated. 4115 * 4116 **********************************************************************/ 4117static int 4118igb_allocate_receive_buffers(struct rx_ring *rxr) 4119{ 4120 struct adapter *adapter = rxr->adapter; 4121 device_t dev = adapter->dev; 4122 struct igb_rx_buf *rxbuf; 4123 int i, bsize, error; 4124 4125 bsize = sizeof(struct igb_rx_buf) * adapter->num_rx_desc; 4126 if (!(rxr->rx_buffers = 4127 (struct igb_rx_buf *) malloc(bsize, 4128 M_DEVBUF, M_NOWAIT | M_ZERO))) { 4129 device_printf(dev, "Unable to allocate rx_buffer memory\n"); 4130 error = ENOMEM; 4131 goto fail; 4132 } 4133 4134 if ((error = bus_dma_tag_create(bus_get_dma_tag(dev), 4135 1, 0, /* alignment, bounds */ 4136 BUS_SPACE_MAXADDR, /* lowaddr */ 4137 BUS_SPACE_MAXADDR, /* highaddr */ 4138 NULL, NULL, /* filter, filterarg */ 4139 MSIZE, /* maxsize */ 4140 1, /* nsegments */ 4141 MSIZE, /* maxsegsize */ 4142 0, /* flags */ 4143 NULL, /* lockfunc */ 4144 NULL, /* lockfuncarg */ 4145 &rxr->htag))) { 4146 device_printf(dev, "Unable to create RX DMA tag\n"); 4147 goto fail; 4148 } 4149 4150 if ((error = bus_dma_tag_create(bus_get_dma_tag(dev), 4151 1, 0, /* alignment, bounds */ 4152 BUS_SPACE_MAXADDR, /* lowaddr */ 4153 BUS_SPACE_MAXADDR, /* highaddr */ 4154 NULL, NULL, /* filter, filterarg */ 4155 MJUM9BYTES, /* maxsize */ 4156 1, /* nsegments */ 4157 MJUM9BYTES, /* maxsegsize */ 4158 0, /* flags */ 4159 NULL, /* lockfunc */ 4160 NULL, /* lockfuncarg */ 4161 &rxr->ptag))) { 4162 device_printf(dev, "Unable to create RX payload DMA tag\n"); 4163 goto fail; 4164 } 4165 4166 for (i = 0; i < adapter->num_rx_desc; i++) { 4167 rxbuf = &rxr->rx_buffers[i]; 4168 error = bus_dmamap_create(rxr->htag, 4169 BUS_DMA_NOWAIT, &rxbuf->hmap); 4170 if (error) { 4171 device_printf(dev, 4172 "Unable to create RX head DMA maps\n"); 4173 goto fail; 4174 } 4175 error = bus_dmamap_create(rxr->ptag, 4176 BUS_DMA_NOWAIT, &rxbuf->pmap); 4177 if (error) { 4178 device_printf(dev, 4179 "Unable to create RX packet DMA maps\n"); 4180 goto fail; 4181 } 4182 } 4183 4184 return (0); 4185 4186fail: 4187 /* Frees all, but can handle partial completion */ 4188 igb_free_receive_structures(adapter); 4189 return (error); 4190} 4191 4192 4193static void 4194igb_free_receive_ring(struct rx_ring *rxr) 4195{ 4196 struct adapter *adapter = rxr->adapter; 4197 struct igb_rx_buf *rxbuf; 4198 4199 4200 for (int i = 0; i < adapter->num_rx_desc; i++) { 4201 rxbuf = &rxr->rx_buffers[i]; 4202 if (rxbuf->m_head != NULL) { 4203 bus_dmamap_sync(rxr->htag, rxbuf->hmap, 4204 BUS_DMASYNC_POSTREAD); 4205 bus_dmamap_unload(rxr->htag, rxbuf->hmap); 4206 rxbuf->m_head->m_flags |= M_PKTHDR; 4207 m_freem(rxbuf->m_head); 4208 } 4209 if (rxbuf->m_pack != NULL) { 4210 bus_dmamap_sync(rxr->ptag, rxbuf->pmap, 4211 BUS_DMASYNC_POSTREAD); 4212 bus_dmamap_unload(rxr->ptag, rxbuf->pmap); 4213 rxbuf->m_pack->m_flags |= M_PKTHDR; 4214 m_freem(rxbuf->m_pack); 4215 } 4216 rxbuf->m_head = NULL; 4217 rxbuf->m_pack = NULL; 4218 } 4219} 4220 4221 4222/********************************************************************* 4223 * 4224 * Initialize a receive ring and its buffers. 4225 * 4226 **********************************************************************/ 4227static int 4228igb_setup_receive_ring(struct rx_ring *rxr) 4229{ 4230 struct adapter *adapter; 4231 struct ifnet *ifp; 4232 device_t dev; 4233 struct igb_rx_buf *rxbuf; 4234 bus_dma_segment_t pseg[1], hseg[1]; 4235 struct lro_ctrl *lro = &rxr->lro; 4236 int rsize, nsegs, error = 0; 4237#ifdef DEV_NETMAP 4238 struct netmap_adapter *na = NA(rxr->adapter->ifp); 4239 struct netmap_slot *slot; 4240#endif /* DEV_NETMAP */ 4241 4242 adapter = rxr->adapter; 4243 dev = adapter->dev; 4244 ifp = adapter->ifp; 4245 4246 /* Clear the ring contents */ 4247 IGB_RX_LOCK(rxr); 4248#ifdef DEV_NETMAP 4249 slot = netmap_reset(na, NR_RX, rxr->me, 0); 4250#endif /* DEV_NETMAP */ 4251 rsize = roundup2(adapter->num_rx_desc * 4252 sizeof(union e1000_adv_rx_desc), IGB_DBA_ALIGN); 4253 bzero((void *)rxr->rx_base, rsize); 4254 4255 /* 4256 ** Free current RX buffer structures and their mbufs 4257 */ 4258 igb_free_receive_ring(rxr); 4259 4260 /* Configure for header split? */ 4261 if (igb_header_split) 4262 rxr->hdr_split = TRUE; 4263 4264 /* Now replenish the ring mbufs */ 4265 for (int j = 0; j < adapter->num_rx_desc; ++j) { 4266 struct mbuf *mh, *mp; 4267 4268 rxbuf = &rxr->rx_buffers[j]; 4269#ifdef DEV_NETMAP 4270 if (slot) { 4271 /* slot sj is mapped to the i-th NIC-ring entry */ 4272 int sj = netmap_idx_n2k(&na->rx_rings[rxr->me], j); 4273 uint64_t paddr; 4274 void *addr; 4275 4276 addr = PNMB(slot + sj, &paddr); 4277 netmap_load_map(rxr->ptag, rxbuf->pmap, addr); 4278 /* Update descriptor */ 4279 rxr->rx_base[j].read.pkt_addr = htole64(paddr); 4280 continue; 4281 } 4282#endif /* DEV_NETMAP */ 4283 if (rxr->hdr_split == FALSE) 4284 goto skip_head; 4285 4286 /* First the header */ 4287 rxbuf->m_head = m_gethdr(M_NOWAIT, MT_DATA); 4288 if (rxbuf->m_head == NULL) { 4289 error = ENOBUFS; 4290 goto fail; 4291 } 4292 m_adj(rxbuf->m_head, ETHER_ALIGN); 4293 mh = rxbuf->m_head; 4294 mh->m_len = mh->m_pkthdr.len = MHLEN; 4295 mh->m_flags |= M_PKTHDR; 4296 /* Get the memory mapping */ 4297 error = bus_dmamap_load_mbuf_sg(rxr->htag, 4298 rxbuf->hmap, rxbuf->m_head, hseg, 4299 &nsegs, BUS_DMA_NOWAIT); 4300 if (error != 0) /* Nothing elegant to do here */ 4301 goto fail; 4302 bus_dmamap_sync(rxr->htag, 4303 rxbuf->hmap, BUS_DMASYNC_PREREAD); 4304 /* Update descriptor */ 4305 rxr->rx_base[j].read.hdr_addr = htole64(hseg[0].ds_addr); 4306 4307skip_head: 4308 /* Now the payload cluster */ 4309 rxbuf->m_pack = m_getjcl(M_NOWAIT, MT_DATA, 4310 M_PKTHDR, adapter->rx_mbuf_sz); 4311 if (rxbuf->m_pack == NULL) { 4312 error = ENOBUFS; 4313 goto fail; 4314 } 4315 mp = rxbuf->m_pack; 4316 mp->m_pkthdr.len = mp->m_len = adapter->rx_mbuf_sz; 4317 /* Get the memory mapping */ 4318 error = bus_dmamap_load_mbuf_sg(rxr->ptag, 4319 rxbuf->pmap, mp, pseg, 4320 &nsegs, BUS_DMA_NOWAIT); 4321 if (error != 0) 4322 goto fail; 4323 bus_dmamap_sync(rxr->ptag, 4324 rxbuf->pmap, BUS_DMASYNC_PREREAD); 4325 /* Update descriptor */ 4326 rxr->rx_base[j].read.pkt_addr = htole64(pseg[0].ds_addr); 4327 } 4328 4329 /* Setup our descriptor indices */ 4330 rxr->next_to_check = 0; 4331 rxr->next_to_refresh = adapter->num_rx_desc - 1; 4332 rxr->lro_enabled = FALSE; 4333 rxr->rx_split_packets = 0; 4334 rxr->rx_bytes = 0; 4335 4336 rxr->fmp = NULL; 4337 rxr->lmp = NULL; 4338 rxr->discard = FALSE; 4339 4340 bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, 4341 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 4342 4343 /* 4344 ** Now set up the LRO interface, we 4345 ** also only do head split when LRO 4346 ** is enabled, since so often they 4347 ** are undesireable in similar setups. 4348 */ 4349 if (ifp->if_capenable & IFCAP_LRO) { 4350 error = tcp_lro_init(lro); 4351 if (error) { 4352 device_printf(dev, "LRO Initialization failed!\n"); 4353 goto fail; 4354 } 4355 INIT_DEBUGOUT("RX LRO Initialized\n"); 4356 rxr->lro_enabled = TRUE; 4357 lro->ifp = adapter->ifp; 4358 } 4359 4360 IGB_RX_UNLOCK(rxr); 4361 return (0); 4362 4363fail: 4364 igb_free_receive_ring(rxr); 4365 IGB_RX_UNLOCK(rxr); 4366 return (error); 4367} 4368 4369 4370/********************************************************************* 4371 * 4372 * Initialize all receive rings. 4373 * 4374 **********************************************************************/ 4375static int 4376igb_setup_receive_structures(struct adapter *adapter) 4377{ 4378 struct rx_ring *rxr = adapter->rx_rings; 4379 int i; 4380 4381 for (i = 0; i < adapter->num_queues; i++, rxr++) 4382 if (igb_setup_receive_ring(rxr)) 4383 goto fail; 4384 4385 return (0); 4386fail: 4387 /* 4388 * Free RX buffers allocated so far, we will only handle 4389 * the rings that completed, the failing case will have 4390 * cleaned up for itself. 'i' is the endpoint. 4391 */ 4392 for (int j = 0; j < i; ++j) { 4393 rxr = &adapter->rx_rings[j]; 4394 IGB_RX_LOCK(rxr); 4395 igb_free_receive_ring(rxr); 4396 IGB_RX_UNLOCK(rxr); 4397 } 4398 4399 return (ENOBUFS); 4400} 4401 4402/********************************************************************* 4403 * 4404 * Enable receive unit. 4405 * 4406 **********************************************************************/ 4407static void 4408igb_initialize_receive_units(struct adapter *adapter) 4409{ 4410 struct rx_ring *rxr = adapter->rx_rings; 4411 struct ifnet *ifp = adapter->ifp; 4412 struct e1000_hw *hw = &adapter->hw; 4413 u32 rctl, rxcsum, psize, srrctl = 0; 4414 4415 INIT_DEBUGOUT("igb_initialize_receive_unit: begin"); 4416 4417 /* 4418 * Make sure receives are disabled while setting 4419 * up the descriptor ring 4420 */ 4421 rctl = E1000_READ_REG(hw, E1000_RCTL); 4422 E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN); 4423 4424 /* 4425 ** Set up for header split 4426 */ 4427 if (igb_header_split) { 4428 /* Use a standard mbuf for the header */ 4429 srrctl |= IGB_HDR_BUF << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT; 4430 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS; 4431 } else 4432 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF; 4433 4434 /* 4435 ** Set up for jumbo frames 4436 */ 4437 if (ifp->if_mtu > ETHERMTU) { 4438 rctl |= E1000_RCTL_LPE; 4439 if (adapter->rx_mbuf_sz == MJUMPAGESIZE) { 4440 srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 4441 rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX; 4442 } else if (adapter->rx_mbuf_sz > MJUMPAGESIZE) { 4443 srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 4444 rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX; 4445 } 4446 /* Set maximum packet len */ 4447 psize = adapter->max_frame_size; 4448 /* are we on a vlan? */ 4449 if (adapter->ifp->if_vlantrunk != NULL) 4450 psize += VLAN_TAG_SIZE; 4451 E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize); 4452 } else { 4453 rctl &= ~E1000_RCTL_LPE; 4454 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 4455 rctl |= E1000_RCTL_SZ_2048; 4456 } 4457 4458 /* Setup the Base and Length of the Rx Descriptor Rings */ 4459 for (int i = 0; i < adapter->num_queues; i++, rxr++) { 4460 u64 bus_addr = rxr->rxdma.dma_paddr; 4461 u32 rxdctl; 4462 4463 E1000_WRITE_REG(hw, E1000_RDLEN(i), 4464 adapter->num_rx_desc * sizeof(struct e1000_rx_desc)); 4465 E1000_WRITE_REG(hw, E1000_RDBAH(i), 4466 (uint32_t)(bus_addr >> 32)); 4467 E1000_WRITE_REG(hw, E1000_RDBAL(i), 4468 (uint32_t)bus_addr); 4469 E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl); 4470 /* Enable this Queue */ 4471 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i)); 4472 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE; 4473 rxdctl &= 0xFFF00000; 4474 rxdctl |= IGB_RX_PTHRESH; 4475 rxdctl |= IGB_RX_HTHRESH << 8; 4476 rxdctl |= IGB_RX_WTHRESH << 16; 4477 E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl); 4478 } 4479 4480 /* 4481 ** Setup for RX MultiQueue 4482 */ 4483 rxcsum = E1000_READ_REG(hw, E1000_RXCSUM); 4484 if (adapter->num_queues >1) { 4485 u32 random[10], mrqc, shift = 0; 4486 union igb_reta { 4487 u32 dword; 4488 u8 bytes[4]; 4489 } reta; 4490 4491 arc4rand(&random, sizeof(random), 0); 4492 if (adapter->hw.mac.type == e1000_82575) 4493 shift = 6; 4494 /* Warning FM follows */ 4495 for (int i = 0; i < 128; i++) { 4496 reta.bytes[i & 3] = 4497 (i % adapter->num_queues) << shift; 4498 if ((i & 3) == 3) 4499 E1000_WRITE_REG(hw, 4500 E1000_RETA(i >> 2), reta.dword); 4501 } 4502 /* Now fill in hash table */ 4503 mrqc = E1000_MRQC_ENABLE_RSS_4Q; 4504 for (int i = 0; i < 10; i++) 4505 E1000_WRITE_REG_ARRAY(hw, 4506 E1000_RSSRK(0), i, random[i]); 4507 4508 mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 | 4509 E1000_MRQC_RSS_FIELD_IPV4_TCP); 4510 mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 | 4511 E1000_MRQC_RSS_FIELD_IPV6_TCP); 4512 mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP | 4513 E1000_MRQC_RSS_FIELD_IPV6_UDP); 4514 mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX | 4515 E1000_MRQC_RSS_FIELD_IPV6_TCP_EX); 4516 4517 E1000_WRITE_REG(hw, E1000_MRQC, mrqc); 4518 4519 /* 4520 ** NOTE: Receive Full-Packet Checksum Offload 4521 ** is mutually exclusive with Multiqueue. However 4522 ** this is not the same as TCP/IP checksums which 4523 ** still work. 4524 */ 4525 rxcsum |= E1000_RXCSUM_PCSD; 4526#if __FreeBSD_version >= 800000 4527 /* For SCTP Offload */ 4528 if ((hw->mac.type == e1000_82576) 4529 && (ifp->if_capenable & IFCAP_RXCSUM)) 4530 rxcsum |= E1000_RXCSUM_CRCOFL; 4531#endif 4532 } else { 4533 /* Non RSS setup */ 4534 if (ifp->if_capenable & IFCAP_RXCSUM) { 4535 rxcsum |= E1000_RXCSUM_IPPCSE; 4536#if __FreeBSD_version >= 800000 4537 if (adapter->hw.mac.type == e1000_82576) 4538 rxcsum |= E1000_RXCSUM_CRCOFL; 4539#endif 4540 } else 4541 rxcsum &= ~E1000_RXCSUM_TUOFL; 4542 } 4543 E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum); 4544 4545 /* Setup the Receive Control Register */ 4546 rctl &= ~(3 << E1000_RCTL_MO_SHIFT); 4547 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO | 4548 E1000_RCTL_RDMTS_HALF | 4549 (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT); 4550 /* Strip CRC bytes. */ 4551 rctl |= E1000_RCTL_SECRC; 4552 /* Make sure VLAN Filters are off */ 4553 rctl &= ~E1000_RCTL_VFE; 4554 /* Don't store bad packets */ 4555 rctl &= ~E1000_RCTL_SBP; 4556 4557 /* Enable Receives */ 4558 E1000_WRITE_REG(hw, E1000_RCTL, rctl); 4559 4560 /* 4561 * Setup the HW Rx Head and Tail Descriptor Pointers 4562 * - needs to be after enable 4563 */ 4564 for (int i = 0; i < adapter->num_queues; i++) { 4565 rxr = &adapter->rx_rings[i]; 4566 E1000_WRITE_REG(hw, E1000_RDH(i), rxr->next_to_check); 4567#ifdef DEV_NETMAP 4568 /* 4569 * an init() while a netmap client is active must 4570 * preserve the rx buffers passed to userspace. 4571 * In this driver it means we adjust RDT to 4572 * somthing different from next_to_refresh 4573 * (which is not used in netmap mode). 4574 */ 4575 if (ifp->if_capenable & IFCAP_NETMAP) { 4576 struct netmap_adapter *na = NA(adapter->ifp); 4577 struct netmap_kring *kring = &na->rx_rings[i]; 4578 int t = rxr->next_to_refresh - kring->nr_hwavail; 4579 4580 if (t >= adapter->num_rx_desc) 4581 t -= adapter->num_rx_desc; 4582 else if (t < 0) 4583 t += adapter->num_rx_desc; 4584 E1000_WRITE_REG(hw, E1000_RDT(i), t); 4585 } else 4586#endif /* DEV_NETMAP */ 4587 E1000_WRITE_REG(hw, E1000_RDT(i), rxr->next_to_refresh); 4588 } 4589 return; 4590} 4591 4592/********************************************************************* 4593 * 4594 * Free receive rings. 4595 * 4596 **********************************************************************/ 4597static void 4598igb_free_receive_structures(struct adapter *adapter) 4599{ 4600 struct rx_ring *rxr = adapter->rx_rings; 4601 4602 for (int i = 0; i < adapter->num_queues; i++, rxr++) { 4603 struct lro_ctrl *lro = &rxr->lro; 4604 igb_free_receive_buffers(rxr); 4605 tcp_lro_free(lro); 4606 igb_dma_free(adapter, &rxr->rxdma); 4607 } 4608 4609 free(adapter->rx_rings, M_DEVBUF); 4610} 4611 4612/********************************************************************* 4613 * 4614 * Free receive ring data structures. 4615 * 4616 **********************************************************************/ 4617static void 4618igb_free_receive_buffers(struct rx_ring *rxr) 4619{ 4620 struct adapter *adapter = rxr->adapter; 4621 struct igb_rx_buf *rxbuf; 4622 int i; 4623 4624 INIT_DEBUGOUT("free_receive_structures: begin"); 4625 4626 /* Cleanup any existing buffers */ 4627 if (rxr->rx_buffers != NULL) { 4628 for (i = 0; i < adapter->num_rx_desc; i++) { 4629 rxbuf = &rxr->rx_buffers[i]; 4630 if (rxbuf->m_head != NULL) { 4631 bus_dmamap_sync(rxr->htag, rxbuf->hmap, 4632 BUS_DMASYNC_POSTREAD); 4633 bus_dmamap_unload(rxr->htag, rxbuf->hmap); 4634 rxbuf->m_head->m_flags |= M_PKTHDR; 4635 m_freem(rxbuf->m_head); 4636 } 4637 if (rxbuf->m_pack != NULL) { 4638 bus_dmamap_sync(rxr->ptag, rxbuf->pmap, 4639 BUS_DMASYNC_POSTREAD); 4640 bus_dmamap_unload(rxr->ptag, rxbuf->pmap); 4641 rxbuf->m_pack->m_flags |= M_PKTHDR; 4642 m_freem(rxbuf->m_pack); 4643 } 4644 rxbuf->m_head = NULL; 4645 rxbuf->m_pack = NULL; 4646 if (rxbuf->hmap != NULL) { 4647 bus_dmamap_destroy(rxr->htag, rxbuf->hmap); 4648 rxbuf->hmap = NULL; 4649 } 4650 if (rxbuf->pmap != NULL) { 4651 bus_dmamap_destroy(rxr->ptag, rxbuf->pmap); 4652 rxbuf->pmap = NULL; 4653 } 4654 } 4655 if (rxr->rx_buffers != NULL) { 4656 free(rxr->rx_buffers, M_DEVBUF); 4657 rxr->rx_buffers = NULL; 4658 } 4659 } 4660 4661 if (rxr->htag != NULL) { 4662 bus_dma_tag_destroy(rxr->htag); 4663 rxr->htag = NULL; 4664 } 4665 if (rxr->ptag != NULL) { 4666 bus_dma_tag_destroy(rxr->ptag); 4667 rxr->ptag = NULL; 4668 } 4669} 4670 4671static __inline void 4672igb_rx_discard(struct rx_ring *rxr, int i) 4673{ 4674 struct igb_rx_buf *rbuf; 4675 4676 rbuf = &rxr->rx_buffers[i]; 4677 4678 /* Partially received? Free the chain */ 4679 if (rxr->fmp != NULL) { 4680 rxr->fmp->m_flags |= M_PKTHDR; 4681 m_freem(rxr->fmp); 4682 rxr->fmp = NULL; 4683 rxr->lmp = NULL; 4684 } 4685 4686 /* 4687 ** With advanced descriptors the writeback 4688 ** clobbers the buffer addrs, so its easier 4689 ** to just free the existing mbufs and take 4690 ** the normal refresh path to get new buffers 4691 ** and mapping. 4692 */ 4693 if (rbuf->m_head) { 4694 m_free(rbuf->m_head); 4695 rbuf->m_head = NULL; 4696 } 4697 4698 if (rbuf->m_pack) { 4699 m_free(rbuf->m_pack); 4700 rbuf->m_pack = NULL; 4701 } 4702 4703 return; 4704} 4705 4706static __inline void 4707igb_rx_input(struct rx_ring *rxr, struct ifnet *ifp, struct mbuf *m, u32 ptype) 4708{ 4709 4710 /* 4711 * ATM LRO is only for IPv4/TCP packets and TCP checksum of the packet 4712 * should be computed by hardware. Also it should not have VLAN tag in 4713 * ethernet header. 4714 */ 4715 if (rxr->lro_enabled && 4716 (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0 && 4717 (ptype & E1000_RXDADV_PKTTYPE_ETQF) == 0 && 4718 (ptype & (E1000_RXDADV_PKTTYPE_IPV4 | E1000_RXDADV_PKTTYPE_TCP)) == 4719 (E1000_RXDADV_PKTTYPE_IPV4 | E1000_RXDADV_PKTTYPE_TCP) && 4720 (m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) == 4721 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) { 4722 /* 4723 * Send to the stack if: 4724 ** - LRO not enabled, or 4725 ** - no LRO resources, or 4726 ** - lro enqueue fails 4727 */ 4728 if (rxr->lro.lro_cnt != 0) 4729 if (tcp_lro_rx(&rxr->lro, m, 0) == 0) 4730 return; 4731 } 4732 IGB_RX_UNLOCK(rxr); 4733 (*ifp->if_input)(ifp, m); 4734 IGB_RX_LOCK(rxr); 4735} 4736 4737/********************************************************************* 4738 * 4739 * This routine executes in interrupt context. It replenishes 4740 * the mbufs in the descriptor and sends data which has been 4741 * dma'ed into host memory to upper layer. 4742 * 4743 * We loop at most count times if count is > 0, or until done if 4744 * count < 0. 4745 * 4746 * Return TRUE if more to clean, FALSE otherwise 4747 *********************************************************************/ 4748static bool 4749igb_rxeof(struct igb_queue *que, int count, int *done) 4750{ 4751 struct adapter *adapter = que->adapter; 4752 struct rx_ring *rxr = que->rxr; 4753 struct ifnet *ifp = adapter->ifp; 4754 struct lro_ctrl *lro = &rxr->lro; 4755 struct lro_entry *queued; 4756 int i, processed = 0, rxdone = 0; 4757 u32 ptype, staterr = 0; 4758 union e1000_adv_rx_desc *cur; 4759 4760 IGB_RX_LOCK(rxr); 4761 /* Sync the ring. */ 4762 bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, 4763 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 4764 4765#ifdef DEV_NETMAP 4766 if (netmap_rx_irq(ifp, rxr->me | NETMAP_LOCKED_ENTER, &processed)) 4767 return (FALSE); 4768#endif /* DEV_NETMAP */ 4769 4770 /* Main clean loop */ 4771 for (i = rxr->next_to_check; count != 0;) { 4772 struct mbuf *sendmp, *mh, *mp; 4773 struct igb_rx_buf *rxbuf; 4774 u16 hlen, plen, hdr, vtag; 4775 bool eop = FALSE; 4776 4777 cur = &rxr->rx_base[i]; 4778 staterr = le32toh(cur->wb.upper.status_error); 4779 if ((staterr & E1000_RXD_STAT_DD) == 0) 4780 break; 4781 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 4782 break; 4783 count--; 4784 sendmp = mh = mp = NULL; 4785 cur->wb.upper.status_error = 0; 4786 rxbuf = &rxr->rx_buffers[i]; 4787 plen = le16toh(cur->wb.upper.length); 4788 ptype = le32toh(cur->wb.lower.lo_dword.data) & IGB_PKTTYPE_MASK; 4789 if ((adapter->hw.mac.type == e1000_i350) && 4790 (staterr & E1000_RXDEXT_STATERR_LB)) 4791 vtag = be16toh(cur->wb.upper.vlan); 4792 else 4793 vtag = le16toh(cur->wb.upper.vlan); 4794 hdr = le16toh(cur->wb.lower.lo_dword.hs_rss.hdr_info); 4795 eop = ((staterr & E1000_RXD_STAT_EOP) == E1000_RXD_STAT_EOP); 4796 4797 /* Make sure all segments of a bad packet are discarded */ 4798 if (((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) != 0) || 4799 (rxr->discard)) { 4800 adapter->dropped_pkts++; 4801 ++rxr->rx_discarded; 4802 if (!eop) /* Catch subsequent segs */ 4803 rxr->discard = TRUE; 4804 else 4805 rxr->discard = FALSE; 4806 igb_rx_discard(rxr, i); 4807 goto next_desc; 4808 } 4809 4810 /* 4811 ** The way the hardware is configured to 4812 ** split, it will ONLY use the header buffer 4813 ** when header split is enabled, otherwise we 4814 ** get normal behavior, ie, both header and 4815 ** payload are DMA'd into the payload buffer. 4816 ** 4817 ** The fmp test is to catch the case where a 4818 ** packet spans multiple descriptors, in that 4819 ** case only the first header is valid. 4820 */ 4821 if (rxr->hdr_split && rxr->fmp == NULL) { 4822 hlen = (hdr & E1000_RXDADV_HDRBUFLEN_MASK) >> 4823 E1000_RXDADV_HDRBUFLEN_SHIFT; 4824 if (hlen > IGB_HDR_BUF) 4825 hlen = IGB_HDR_BUF; 4826 mh = rxr->rx_buffers[i].m_head; 4827 mh->m_len = hlen; 4828 /* clear buf pointer for refresh */ 4829 rxbuf->m_head = NULL; 4830 /* 4831 ** Get the payload length, this 4832 ** could be zero if its a small 4833 ** packet. 4834 */ 4835 if (plen > 0) { 4836 mp = rxr->rx_buffers[i].m_pack; 4837 mp->m_len = plen; 4838 mh->m_next = mp; 4839 /* clear buf pointer */ 4840 rxbuf->m_pack = NULL; 4841 rxr->rx_split_packets++; 4842 } 4843 } else { 4844 /* 4845 ** Either no header split, or a 4846 ** secondary piece of a fragmented 4847 ** split packet. 4848 */ 4849 mh = rxr->rx_buffers[i].m_pack; 4850 mh->m_len = plen; 4851 /* clear buf info for refresh */ 4852 rxbuf->m_pack = NULL; 4853 } 4854 4855 ++processed; /* So we know when to refresh */ 4856 4857 /* Initial frame - setup */ 4858 if (rxr->fmp == NULL) { 4859 mh->m_pkthdr.len = mh->m_len; 4860 /* Save the head of the chain */ 4861 rxr->fmp = mh; 4862 rxr->lmp = mh; 4863 if (mp != NULL) { 4864 /* Add payload if split */ 4865 mh->m_pkthdr.len += mp->m_len; 4866 rxr->lmp = mh->m_next; 4867 } 4868 } else { 4869 /* Chain mbuf's together */ 4870 rxr->lmp->m_next = mh; 4871 rxr->lmp = rxr->lmp->m_next; 4872 rxr->fmp->m_pkthdr.len += mh->m_len; 4873 } 4874 4875 if (eop) { 4876 rxr->fmp->m_pkthdr.rcvif = ifp; 4877 ifp->if_ipackets++; 4878 rxr->rx_packets++; 4879 /* capture data for AIM */ 4880 rxr->packets++; 4881 rxr->bytes += rxr->fmp->m_pkthdr.len; 4882 rxr->rx_bytes += rxr->fmp->m_pkthdr.len; 4883 4884 if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) 4885 igb_rx_checksum(staterr, rxr->fmp, ptype); 4886 4887 if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0 && 4888 (staterr & E1000_RXD_STAT_VP) != 0) { 4889 rxr->fmp->m_pkthdr.ether_vtag = vtag; 4890 rxr->fmp->m_flags |= M_VLANTAG; 4891 } 4892#ifndef IGB_LEGACY_TX 4893 rxr->fmp->m_pkthdr.flowid = que->msix; 4894 rxr->fmp->m_flags |= M_FLOWID; 4895#endif 4896 sendmp = rxr->fmp; 4897 /* Make sure to set M_PKTHDR. */ 4898 sendmp->m_flags |= M_PKTHDR; 4899 rxr->fmp = NULL; 4900 rxr->lmp = NULL; 4901 } 4902 4903next_desc: 4904 bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, 4905 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 4906 4907 /* Advance our pointers to the next descriptor. */ 4908 if (++i == adapter->num_rx_desc) 4909 i = 0; 4910 /* 4911 ** Send to the stack or LRO 4912 */ 4913 if (sendmp != NULL) { 4914 rxr->next_to_check = i; 4915 igb_rx_input(rxr, ifp, sendmp, ptype); 4916 i = rxr->next_to_check; 4917 rxdone++; 4918 } 4919 4920 /* Every 8 descriptors we go to refresh mbufs */ 4921 if (processed == 8) { 4922 igb_refresh_mbufs(rxr, i); 4923 processed = 0; 4924 } 4925 } 4926 4927 /* Catch any remainders */ 4928 if (igb_rx_unrefreshed(rxr)) 4929 igb_refresh_mbufs(rxr, i); 4930 4931 rxr->next_to_check = i; 4932 4933 /* 4934 * Flush any outstanding LRO work 4935 */ 4936 while ((queued = SLIST_FIRST(&lro->lro_active)) != NULL) { 4937 SLIST_REMOVE_HEAD(&lro->lro_active, next); 4938 tcp_lro_flush(lro, queued); 4939 } 4940 4941 if (done != NULL) 4942 *done += rxdone; 4943 4944 IGB_RX_UNLOCK(rxr); 4945 return ((staterr & E1000_RXD_STAT_DD) ? TRUE : FALSE); 4946} 4947 4948/********************************************************************* 4949 * 4950 * Verify that the hardware indicated that the checksum is valid. 4951 * Inform the stack about the status of checksum so that stack 4952 * doesn't spend time verifying the checksum. 4953 * 4954 *********************************************************************/ 4955static void 4956igb_rx_checksum(u32 staterr, struct mbuf *mp, u32 ptype) 4957{ 4958 u16 status = (u16)staterr; 4959 u8 errors = (u8) (staterr >> 24); 4960 int sctp; 4961 4962 /* Ignore Checksum bit is set */ 4963 if (status & E1000_RXD_STAT_IXSM) { 4964 mp->m_pkthdr.csum_flags = 0; 4965 return; 4966 } 4967 4968 if ((ptype & E1000_RXDADV_PKTTYPE_ETQF) == 0 && 4969 (ptype & E1000_RXDADV_PKTTYPE_SCTP) != 0) 4970 sctp = 1; 4971 else 4972 sctp = 0; 4973 if (status & E1000_RXD_STAT_IPCS) { 4974 /* Did it pass? */ 4975 if (!(errors & E1000_RXD_ERR_IPE)) { 4976 /* IP Checksum Good */ 4977 mp->m_pkthdr.csum_flags = CSUM_IP_CHECKED; 4978 mp->m_pkthdr.csum_flags |= CSUM_IP_VALID; 4979 } else 4980 mp->m_pkthdr.csum_flags = 0; 4981 } 4982 4983 if (status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)) { 4984 u16 type = (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 4985#if __FreeBSD_version >= 800000 4986 if (sctp) /* reassign */ 4987 type = CSUM_SCTP_VALID; 4988#endif 4989 /* Did it pass? */ 4990 if (!(errors & E1000_RXD_ERR_TCPE)) { 4991 mp->m_pkthdr.csum_flags |= type; 4992 if (sctp == 0) 4993 mp->m_pkthdr.csum_data = htons(0xffff); 4994 } 4995 } 4996 return; 4997} 4998 4999/* 5000 * This routine is run via an vlan 5001 * config EVENT 5002 */ 5003static void 5004igb_register_vlan(void *arg, struct ifnet *ifp, u16 vtag) 5005{ 5006 struct adapter *adapter = ifp->if_softc; 5007 u32 index, bit; 5008 5009 if (ifp->if_softc != arg) /* Not our event */ 5010 return; 5011 5012 if ((vtag == 0) || (vtag > 4095)) /* Invalid */ 5013 return; 5014 5015 IGB_CORE_LOCK(adapter); 5016 index = (vtag >> 5) & 0x7F; 5017 bit = vtag & 0x1F; 5018 adapter->shadow_vfta[index] |= (1 << bit); 5019 ++adapter->num_vlans; 5020 /* Change hw filter setting */ 5021 if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) 5022 igb_setup_vlan_hw_support(adapter); 5023 IGB_CORE_UNLOCK(adapter); 5024} 5025 5026/* 5027 * This routine is run via an vlan 5028 * unconfig EVENT 5029 */ 5030static void 5031igb_unregister_vlan(void *arg, struct ifnet *ifp, u16 vtag) 5032{ 5033 struct adapter *adapter = ifp->if_softc; 5034 u32 index, bit; 5035 5036 if (ifp->if_softc != arg) 5037 return; 5038 5039 if ((vtag == 0) || (vtag > 4095)) /* Invalid */ 5040 return; 5041 5042 IGB_CORE_LOCK(adapter); 5043 index = (vtag >> 5) & 0x7F; 5044 bit = vtag & 0x1F; 5045 adapter->shadow_vfta[index] &= ~(1 << bit); 5046 --adapter->num_vlans; 5047 /* Change hw filter setting */ 5048 if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) 5049 igb_setup_vlan_hw_support(adapter); 5050 IGB_CORE_UNLOCK(adapter); 5051} 5052 5053static void 5054igb_setup_vlan_hw_support(struct adapter *adapter) 5055{ 5056 struct e1000_hw *hw = &adapter->hw; 5057 struct ifnet *ifp = adapter->ifp; 5058 u32 reg; 5059 5060 if (adapter->vf_ifp) { 5061 e1000_rlpml_set_vf(hw, 5062 adapter->max_frame_size + VLAN_TAG_SIZE); 5063 return; 5064 } 5065 5066 reg = E1000_READ_REG(hw, E1000_CTRL); 5067 reg |= E1000_CTRL_VME; 5068 E1000_WRITE_REG(hw, E1000_CTRL, reg); 5069 5070 /* Enable the Filter Table */ 5071 if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) { 5072 reg = E1000_READ_REG(hw, E1000_RCTL); 5073 reg &= ~E1000_RCTL_CFIEN; 5074 reg |= E1000_RCTL_VFE; 5075 E1000_WRITE_REG(hw, E1000_RCTL, reg); 5076 } 5077 5078 /* Update the frame size */ 5079 E1000_WRITE_REG(&adapter->hw, E1000_RLPML, 5080 adapter->max_frame_size + VLAN_TAG_SIZE); 5081 5082 /* Don't bother with table if no vlans */ 5083 if ((adapter->num_vlans == 0) || 5084 ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0)) 5085 return; 5086 /* 5087 ** A soft reset zero's out the VFTA, so 5088 ** we need to repopulate it now. 5089 */ 5090 for (int i = 0; i < IGB_VFTA_SIZE; i++) 5091 if (adapter->shadow_vfta[i] != 0) { 5092 if (adapter->vf_ifp) 5093 e1000_vfta_set_vf(hw, 5094 adapter->shadow_vfta[i], TRUE); 5095 else 5096 e1000_write_vfta(hw, 5097 i, adapter->shadow_vfta[i]); 5098 } 5099} 5100 5101static void 5102igb_enable_intr(struct adapter *adapter) 5103{ 5104 /* With RSS set up what to auto clear */ 5105 if (adapter->msix_mem) { 5106 u32 mask = (adapter->que_mask | adapter->link_mask); 5107 E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask); 5108 E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask); 5109 E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask); 5110 E1000_WRITE_REG(&adapter->hw, E1000_IMS, 5111 E1000_IMS_LSC); 5112 } else { 5113 E1000_WRITE_REG(&adapter->hw, E1000_IMS, 5114 IMS_ENABLE_MASK); 5115 } 5116 E1000_WRITE_FLUSH(&adapter->hw); 5117 5118 return; 5119} 5120 5121static void 5122igb_disable_intr(struct adapter *adapter) 5123{ 5124 if (adapter->msix_mem) { 5125 E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0); 5126 E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0); 5127 } 5128 E1000_WRITE_REG(&adapter->hw, E1000_IMC, ~0); 5129 E1000_WRITE_FLUSH(&adapter->hw); 5130 return; 5131} 5132 5133/* 5134 * Bit of a misnomer, what this really means is 5135 * to enable OS management of the system... aka 5136 * to disable special hardware management features 5137 */ 5138static void 5139igb_init_manageability(struct adapter *adapter) 5140{ 5141 if (adapter->has_manage) { 5142 int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H); 5143 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC); 5144 5145 /* disable hardware interception of ARP */ 5146 manc &= ~(E1000_MANC_ARP_EN); 5147 5148 /* enable receiving management packets to the host */ 5149 manc |= E1000_MANC_EN_MNG2HOST; 5150 manc2h |= 1 << 5; /* Mng Port 623 */ 5151 manc2h |= 1 << 6; /* Mng Port 664 */ 5152 E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h); 5153 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc); 5154 } 5155} 5156 5157/* 5158 * Give control back to hardware management 5159 * controller if there is one. 5160 */ 5161static void 5162igb_release_manageability(struct adapter *adapter) 5163{ 5164 if (adapter->has_manage) { 5165 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC); 5166 5167 /* re-enable hardware interception of ARP */ 5168 manc |= E1000_MANC_ARP_EN; 5169 manc &= ~E1000_MANC_EN_MNG2HOST; 5170 5171 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc); 5172 } 5173} 5174 5175/* 5176 * igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit. 5177 * For ASF and Pass Through versions of f/w this means that 5178 * the driver is loaded. 5179 * 5180 */ 5181static void 5182igb_get_hw_control(struct adapter *adapter) 5183{ 5184 u32 ctrl_ext; 5185 5186 if (adapter->vf_ifp) 5187 return; 5188 5189 /* Let firmware know the driver has taken over */ 5190 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT); 5191 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, 5192 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD); 5193} 5194 5195/* 5196 * igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit. 5197 * For ASF and Pass Through versions of f/w this means that the 5198 * driver is no longer loaded. 5199 * 5200 */ 5201static void 5202igb_release_hw_control(struct adapter *adapter) 5203{ 5204 u32 ctrl_ext; 5205 5206 if (adapter->vf_ifp) 5207 return; 5208 5209 /* Let firmware taken over control of h/w */ 5210 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT); 5211 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, 5212 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD); 5213} 5214 5215static int 5216igb_is_valid_ether_addr(uint8_t *addr) 5217{ 5218 char zero_addr[6] = { 0, 0, 0, 0, 0, 0 }; 5219 5220 if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) { 5221 return (FALSE); 5222 } 5223 5224 return (TRUE); 5225} 5226 5227 5228/* 5229 * Enable PCI Wake On Lan capability 5230 */ 5231static void 5232igb_enable_wakeup(device_t dev) 5233{ 5234 u16 cap, status; 5235 u8 id; 5236 5237 /* First find the capabilities pointer*/ 5238 cap = pci_read_config(dev, PCIR_CAP_PTR, 2); 5239 /* Read the PM Capabilities */ 5240 id = pci_read_config(dev, cap, 1); 5241 if (id != PCIY_PMG) /* Something wrong */ 5242 return; 5243 /* OK, we have the power capabilities, so 5244 now get the status register */ 5245 cap += PCIR_POWER_STATUS; 5246 status = pci_read_config(dev, cap, 2); 5247 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE; 5248 pci_write_config(dev, cap, status, 2); 5249 return; 5250} 5251 5252static void 5253igb_led_func(void *arg, int onoff) 5254{ 5255 struct adapter *adapter = arg; 5256 5257 IGB_CORE_LOCK(adapter); 5258 if (onoff) { 5259 e1000_setup_led(&adapter->hw); 5260 e1000_led_on(&adapter->hw); 5261 } else { 5262 e1000_led_off(&adapter->hw); 5263 e1000_cleanup_led(&adapter->hw); 5264 } 5265 IGB_CORE_UNLOCK(adapter); 5266} 5267 5268/********************************************************************** 5269 * 5270 * Update the board statistics counters. 5271 * 5272 **********************************************************************/ 5273static void 5274igb_update_stats_counters(struct adapter *adapter) 5275{ 5276 struct ifnet *ifp; 5277 struct e1000_hw *hw = &adapter->hw; 5278 struct e1000_hw_stats *stats; 5279 5280 /* 5281 ** The virtual function adapter has only a 5282 ** small controlled set of stats, do only 5283 ** those and return. 5284 */ 5285 if (adapter->vf_ifp) { 5286 igb_update_vf_stats_counters(adapter); 5287 return; 5288 } 5289 5290 stats = (struct e1000_hw_stats *)adapter->stats; 5291 5292 if(adapter->hw.phy.media_type == e1000_media_type_copper || 5293 (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) { 5294 stats->symerrs += 5295 E1000_READ_REG(hw,E1000_SYMERRS); 5296 stats->sec += E1000_READ_REG(hw, E1000_SEC); 5297 } 5298 5299 stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS); 5300 stats->mpc += E1000_READ_REG(hw, E1000_MPC); 5301 stats->scc += E1000_READ_REG(hw, E1000_SCC); 5302 stats->ecol += E1000_READ_REG(hw, E1000_ECOL); 5303 5304 stats->mcc += E1000_READ_REG(hw, E1000_MCC); 5305 stats->latecol += E1000_READ_REG(hw, E1000_LATECOL); 5306 stats->colc += E1000_READ_REG(hw, E1000_COLC); 5307 stats->dc += E1000_READ_REG(hw, E1000_DC); 5308 stats->rlec += E1000_READ_REG(hw, E1000_RLEC); 5309 stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC); 5310 stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC); 5311 /* 5312 ** For watchdog management we need to know if we have been 5313 ** paused during the last interval, so capture that here. 5314 */ 5315 adapter->pause_frames = E1000_READ_REG(&adapter->hw, E1000_XOFFRXC); 5316 stats->xoffrxc += adapter->pause_frames; 5317 stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC); 5318 stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC); 5319 stats->prc64 += E1000_READ_REG(hw, E1000_PRC64); 5320 stats->prc127 += E1000_READ_REG(hw, E1000_PRC127); 5321 stats->prc255 += E1000_READ_REG(hw, E1000_PRC255); 5322 stats->prc511 += E1000_READ_REG(hw, E1000_PRC511); 5323 stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023); 5324 stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522); 5325 stats->gprc += E1000_READ_REG(hw, E1000_GPRC); 5326 stats->bprc += E1000_READ_REG(hw, E1000_BPRC); 5327 stats->mprc += E1000_READ_REG(hw, E1000_MPRC); 5328 stats->gptc += E1000_READ_REG(hw, E1000_GPTC); 5329 5330 /* For the 64-bit byte counters the low dword must be read first. */ 5331 /* Both registers clear on the read of the high dword */ 5332 5333 stats->gorc += E1000_READ_REG(hw, E1000_GORCL) + 5334 ((u64)E1000_READ_REG(hw, E1000_GORCH) << 32); 5335 stats->gotc += E1000_READ_REG(hw, E1000_GOTCL) + 5336 ((u64)E1000_READ_REG(hw, E1000_GOTCH) << 32); 5337 5338 stats->rnbc += E1000_READ_REG(hw, E1000_RNBC); 5339 stats->ruc += E1000_READ_REG(hw, E1000_RUC); 5340 stats->rfc += E1000_READ_REG(hw, E1000_RFC); 5341 stats->roc += E1000_READ_REG(hw, E1000_ROC); 5342 stats->rjc += E1000_READ_REG(hw, E1000_RJC); 5343 5344 stats->tor += E1000_READ_REG(hw, E1000_TORH); 5345 stats->tot += E1000_READ_REG(hw, E1000_TOTH); 5346 5347 stats->tpr += E1000_READ_REG(hw, E1000_TPR); 5348 stats->tpt += E1000_READ_REG(hw, E1000_TPT); 5349 stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64); 5350 stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127); 5351 stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255); 5352 stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511); 5353 stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023); 5354 stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522); 5355 stats->mptc += E1000_READ_REG(hw, E1000_MPTC); 5356 stats->bptc += E1000_READ_REG(hw, E1000_BPTC); 5357 5358 /* Interrupt Counts */ 5359 5360 stats->iac += E1000_READ_REG(hw, E1000_IAC); 5361 stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC); 5362 stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC); 5363 stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC); 5364 stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC); 5365 stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC); 5366 stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC); 5367 stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC); 5368 stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC); 5369 5370 /* Host to Card Statistics */ 5371 5372 stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC); 5373 stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC); 5374 stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC); 5375 stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC); 5376 stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC); 5377 stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC); 5378 stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC); 5379 stats->hgorc += (E1000_READ_REG(hw, E1000_HGORCL) + 5380 ((u64)E1000_READ_REG(hw, E1000_HGORCH) << 32)); 5381 stats->hgotc += (E1000_READ_REG(hw, E1000_HGOTCL) + 5382 ((u64)E1000_READ_REG(hw, E1000_HGOTCH) << 32)); 5383 stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS); 5384 stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC); 5385 stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC); 5386 5387 stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC); 5388 stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC); 5389 stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS); 5390 stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR); 5391 stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC); 5392 stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC); 5393 5394 ifp = adapter->ifp; 5395 ifp->if_collisions = stats->colc; 5396 5397 /* Rx Errors */ 5398 ifp->if_ierrors = adapter->dropped_pkts + stats->rxerrc + 5399 stats->crcerrs + stats->algnerrc + 5400 stats->ruc + stats->roc + stats->mpc + stats->cexterr; 5401 5402 /* Tx Errors */ 5403 ifp->if_oerrors = stats->ecol + 5404 stats->latecol + adapter->watchdog_events; 5405 5406 /* Driver specific counters */ 5407 adapter->device_control = E1000_READ_REG(hw, E1000_CTRL); 5408 adapter->rx_control = E1000_READ_REG(hw, E1000_RCTL); 5409 adapter->int_mask = E1000_READ_REG(hw, E1000_IMS); 5410 adapter->eint_mask = E1000_READ_REG(hw, E1000_EIMS); 5411 adapter->packet_buf_alloc_tx = 5412 ((E1000_READ_REG(hw, E1000_PBA) & 0xffff0000) >> 16); 5413 adapter->packet_buf_alloc_rx = 5414 (E1000_READ_REG(hw, E1000_PBA) & 0xffff); 5415} 5416 5417 5418/********************************************************************** 5419 * 5420 * Initialize the VF board statistics counters. 5421 * 5422 **********************************************************************/ 5423static void 5424igb_vf_init_stats(struct adapter *adapter) 5425{ 5426 struct e1000_hw *hw = &adapter->hw; 5427 struct e1000_vf_stats *stats; 5428 5429 stats = (struct e1000_vf_stats *)adapter->stats; 5430 if (stats == NULL) 5431 return; 5432 stats->last_gprc = E1000_READ_REG(hw, E1000_VFGPRC); 5433 stats->last_gorc = E1000_READ_REG(hw, E1000_VFGORC); 5434 stats->last_gptc = E1000_READ_REG(hw, E1000_VFGPTC); 5435 stats->last_gotc = E1000_READ_REG(hw, E1000_VFGOTC); 5436 stats->last_mprc = E1000_READ_REG(hw, E1000_VFMPRC); 5437} 5438 5439/********************************************************************** 5440 * 5441 * Update the VF board statistics counters. 5442 * 5443 **********************************************************************/ 5444static void 5445igb_update_vf_stats_counters(struct adapter *adapter) 5446{ 5447 struct e1000_hw *hw = &adapter->hw; 5448 struct e1000_vf_stats *stats; 5449 5450 if (adapter->link_speed == 0) 5451 return; 5452 5453 stats = (struct e1000_vf_stats *)adapter->stats; 5454 5455 UPDATE_VF_REG(E1000_VFGPRC, 5456 stats->last_gprc, stats->gprc); 5457 UPDATE_VF_REG(E1000_VFGORC, 5458 stats->last_gorc, stats->gorc); 5459 UPDATE_VF_REG(E1000_VFGPTC, 5460 stats->last_gptc, stats->gptc); 5461 UPDATE_VF_REG(E1000_VFGOTC, 5462 stats->last_gotc, stats->gotc); 5463 UPDATE_VF_REG(E1000_VFMPRC, 5464 stats->last_mprc, stats->mprc); 5465} 5466 5467/* Export a single 32-bit register via a read-only sysctl. */ 5468static int 5469igb_sysctl_reg_handler(SYSCTL_HANDLER_ARGS) 5470{ 5471 struct adapter *adapter; 5472 u_int val; 5473 5474 adapter = oidp->oid_arg1; 5475 val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2); 5476 return (sysctl_handle_int(oidp, &val, 0, req)); 5477} 5478 5479/* 5480** Tuneable interrupt rate handler 5481*/ 5482static int 5483igb_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS) 5484{ 5485 struct igb_queue *que = ((struct igb_queue *)oidp->oid_arg1); 5486 int error; 5487 u32 reg, usec, rate; 5488 5489 reg = E1000_READ_REG(&que->adapter->hw, E1000_EITR(que->msix)); 5490 usec = ((reg & 0x7FFC) >> 2); 5491 if (usec > 0) 5492 rate = 1000000 / usec; 5493 else 5494 rate = 0; 5495 error = sysctl_handle_int(oidp, &rate, 0, req); 5496 if (error || !req->newptr) 5497 return error; 5498 return 0; 5499} 5500 5501/* 5502 * Add sysctl variables, one per statistic, to the system. 5503 */ 5504static void 5505igb_add_hw_stats(struct adapter *adapter) 5506{ 5507 device_t dev = adapter->dev; 5508 5509 struct tx_ring *txr = adapter->tx_rings; 5510 struct rx_ring *rxr = adapter->rx_rings; 5511 5512 struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); 5513 struct sysctl_oid *tree = device_get_sysctl_tree(dev); 5514 struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree); 5515 struct e1000_hw_stats *stats = adapter->stats; 5516 5517 struct sysctl_oid *stat_node, *queue_node, *int_node, *host_node; 5518 struct sysctl_oid_list *stat_list, *queue_list, *int_list, *host_list; 5519 5520#define QUEUE_NAME_LEN 32 5521 char namebuf[QUEUE_NAME_LEN]; 5522 5523 /* Driver Statistics */ 5524 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "link_irq", 5525 CTLFLAG_RD, &adapter->link_irq, 0, 5526 "Link MSIX IRQ Handled"); 5527 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped", 5528 CTLFLAG_RD, &adapter->dropped_pkts, 5529 "Driver dropped packets"); 5530 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail", 5531 CTLFLAG_RD, &adapter->no_tx_dma_setup, 5532 "Driver tx dma failure in xmit"); 5533 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns", 5534 CTLFLAG_RD, &adapter->rx_overruns, 5535 "RX overruns"); 5536 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts", 5537 CTLFLAG_RD, &adapter->watchdog_events, 5538 "Watchdog timeouts"); 5539 5540 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "device_control", 5541 CTLFLAG_RD, &adapter->device_control, 5542 "Device Control Register"); 5543 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_control", 5544 CTLFLAG_RD, &adapter->rx_control, 5545 "Receiver Control Register"); 5546 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "interrupt_mask", 5547 CTLFLAG_RD, &adapter->int_mask, 5548 "Interrupt Mask"); 5549 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "extended_int_mask", 5550 CTLFLAG_RD, &adapter->eint_mask, 5551 "Extended Interrupt Mask"); 5552 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_buf_alloc", 5553 CTLFLAG_RD, &adapter->packet_buf_alloc_tx, 5554 "Transmit Buffer Packet Allocation"); 5555 SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_buf_alloc", 5556 CTLFLAG_RD, &adapter->packet_buf_alloc_rx, 5557 "Receive Buffer Packet Allocation"); 5558 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water", 5559 CTLFLAG_RD, &adapter->hw.fc.high_water, 0, 5560 "Flow Control High Watermark"); 5561 SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water", 5562 CTLFLAG_RD, &adapter->hw.fc.low_water, 0, 5563 "Flow Control Low Watermark"); 5564 5565 for (int i = 0; i < adapter->num_queues; i++, rxr++, txr++) { 5566 struct lro_ctrl *lro = &rxr->lro; 5567 5568 snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); 5569 queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, 5570 CTLFLAG_RD, NULL, "Queue Name"); 5571 queue_list = SYSCTL_CHILDREN(queue_node); 5572 5573 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate", 5574 CTLFLAG_RD, &adapter->queues[i], 5575 sizeof(&adapter->queues[i]), 5576 igb_sysctl_interrupt_rate_handler, 5577 "IU", "Interrupt Rate"); 5578 5579 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head", 5580 CTLFLAG_RD, adapter, E1000_TDH(txr->me), 5581 igb_sysctl_reg_handler, "IU", 5582 "Transmit Descriptor Head"); 5583 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail", 5584 CTLFLAG_RD, adapter, E1000_TDT(txr->me), 5585 igb_sysctl_reg_handler, "IU", 5586 "Transmit Descriptor Tail"); 5587 SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "no_desc_avail", 5588 CTLFLAG_RD, &txr->no_desc_avail, 5589 "Queue No Descriptor Available"); 5590 SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "tx_packets", 5591 CTLFLAG_RD, &txr->tx_packets, 5592 "Queue Packets Transmitted"); 5593 5594 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head", 5595 CTLFLAG_RD, adapter, E1000_RDH(rxr->me), 5596 igb_sysctl_reg_handler, "IU", 5597 "Receive Descriptor Head"); 5598 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail", 5599 CTLFLAG_RD, adapter, E1000_RDT(rxr->me), 5600 igb_sysctl_reg_handler, "IU", 5601 "Receive Descriptor Tail"); 5602 SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "rx_packets", 5603 CTLFLAG_RD, &rxr->rx_packets, 5604 "Queue Packets Received"); 5605 SYSCTL_ADD_QUAD(ctx, queue_list, OID_AUTO, "rx_bytes", 5606 CTLFLAG_RD, &rxr->rx_bytes, 5607 "Queue Bytes Received"); 5608 SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "lro_queued", 5609 CTLFLAG_RD, &lro->lro_queued, 0, 5610 "LRO Queued"); 5611 SYSCTL_ADD_UINT(ctx, queue_list, OID_AUTO, "lro_flushed", 5612 CTLFLAG_RD, &lro->lro_flushed, 0, 5613 "LRO Flushed"); 5614 } 5615 5616 /* MAC stats get their own sub node */ 5617 5618 stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats", 5619 CTLFLAG_RD, NULL, "MAC Statistics"); 5620 stat_list = SYSCTL_CHILDREN(stat_node); 5621 5622 /* 5623 ** VF adapter has a very limited set of stats 5624 ** since its not managing the metal, so to speak. 5625 */ 5626 if (adapter->vf_ifp) { 5627 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd", 5628 CTLFLAG_RD, &stats->gprc, 5629 "Good Packets Received"); 5630 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd", 5631 CTLFLAG_RD, &stats->gptc, 5632 "Good Packets Transmitted"); 5633 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd", 5634 CTLFLAG_RD, &stats->gorc, 5635 "Good Octets Received"); 5636 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_octets_txd", 5637 CTLFLAG_RD, &stats->gotc, 5638 "Good Octets Transmitted"); 5639 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd", 5640 CTLFLAG_RD, &stats->mprc, 5641 "Multicast Packets Received"); 5642 return; 5643 } 5644 5645 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "excess_coll", 5646 CTLFLAG_RD, &stats->ecol, 5647 "Excessive collisions"); 5648 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "single_coll", 5649 CTLFLAG_RD, &stats->scc, 5650 "Single collisions"); 5651 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "multiple_coll", 5652 CTLFLAG_RD, &stats->mcc, 5653 "Multiple collisions"); 5654 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "late_coll", 5655 CTLFLAG_RD, &stats->latecol, 5656 "Late collisions"); 5657 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "collision_count", 5658 CTLFLAG_RD, &stats->colc, 5659 "Collision Count"); 5660 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "symbol_errors", 5661 CTLFLAG_RD, &stats->symerrs, 5662 "Symbol Errors"); 5663 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "sequence_errors", 5664 CTLFLAG_RD, &stats->sec, 5665 "Sequence Errors"); 5666 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "defer_count", 5667 CTLFLAG_RD, &stats->dc, 5668 "Defer Count"); 5669 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "missed_packets", 5670 CTLFLAG_RD, &stats->mpc, 5671 "Missed Packets"); 5672 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "recv_no_buff", 5673 CTLFLAG_RD, &stats->rnbc, 5674 "Receive No Buffers"); 5675 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "recv_undersize", 5676 CTLFLAG_RD, &stats->ruc, 5677 "Receive Undersize"); 5678 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "recv_fragmented", 5679 CTLFLAG_RD, &stats->rfc, 5680 "Fragmented Packets Received "); 5681 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "recv_oversize", 5682 CTLFLAG_RD, &stats->roc, 5683 "Oversized Packets Received"); 5684 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "recv_jabber", 5685 CTLFLAG_RD, &stats->rjc, 5686 "Recevied Jabber"); 5687 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "recv_errs", 5688 CTLFLAG_RD, &stats->rxerrc, 5689 "Receive Errors"); 5690 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "crc_errs", 5691 CTLFLAG_RD, &stats->crcerrs, 5692 "CRC errors"); 5693 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "alignment_errs", 5694 CTLFLAG_RD, &stats->algnerrc, 5695 "Alignment Errors"); 5696 /* On 82575 these are collision counts */ 5697 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs", 5698 CTLFLAG_RD, &stats->cexterr, 5699 "Collision/Carrier extension errors"); 5700 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "xon_recvd", 5701 CTLFLAG_RD, &stats->xonrxc, 5702 "XON Received"); 5703 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "xon_txd", 5704 CTLFLAG_RD, &stats->xontxc, 5705 "XON Transmitted"); 5706 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "xoff_recvd", 5707 CTLFLAG_RD, &stats->xoffrxc, 5708 "XOFF Received"); 5709 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "xoff_txd", 5710 CTLFLAG_RD, &stats->xofftxc, 5711 "XOFF Transmitted"); 5712 /* Packet Reception Stats */ 5713 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd", 5714 CTLFLAG_RD, &stats->tpr, 5715 "Total Packets Received "); 5716 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd", 5717 CTLFLAG_RD, &stats->gprc, 5718 "Good Packets Received"); 5719 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd", 5720 CTLFLAG_RD, &stats->bprc, 5721 "Broadcast Packets Received"); 5722 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd", 5723 CTLFLAG_RD, &stats->mprc, 5724 "Multicast Packets Received"); 5725 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "rx_frames_64", 5726 CTLFLAG_RD, &stats->prc64, 5727 "64 byte frames received "); 5728 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127", 5729 CTLFLAG_RD, &stats->prc127, 5730 "65-127 byte frames received"); 5731 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255", 5732 CTLFLAG_RD, &stats->prc255, 5733 "128-255 byte frames received"); 5734 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511", 5735 CTLFLAG_RD, &stats->prc511, 5736 "256-511 byte frames received"); 5737 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023", 5738 CTLFLAG_RD, &stats->prc1023, 5739 "512-1023 byte frames received"); 5740 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522", 5741 CTLFLAG_RD, &stats->prc1522, 5742 "1023-1522 byte frames received"); 5743 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd", 5744 CTLFLAG_RD, &stats->gorc, 5745 "Good Octets Received"); 5746 5747 /* Packet Transmission Stats */ 5748 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_octets_txd", 5749 CTLFLAG_RD, &stats->gotc, 5750 "Good Octets Transmitted"); 5751 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd", 5752 CTLFLAG_RD, &stats->tpt, 5753 "Total Packets Transmitted"); 5754 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd", 5755 CTLFLAG_RD, &stats->gptc, 5756 "Good Packets Transmitted"); 5757 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd", 5758 CTLFLAG_RD, &stats->bptc, 5759 "Broadcast Packets Transmitted"); 5760 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd", 5761 CTLFLAG_RD, &stats->mptc, 5762 "Multicast Packets Transmitted"); 5763 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tx_frames_64", 5764 CTLFLAG_RD, &stats->ptc64, 5765 "64 byte frames transmitted "); 5766 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127", 5767 CTLFLAG_RD, &stats->ptc127, 5768 "65-127 byte frames transmitted"); 5769 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255", 5770 CTLFLAG_RD, &stats->ptc255, 5771 "128-255 byte frames transmitted"); 5772 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511", 5773 CTLFLAG_RD, &stats->ptc511, 5774 "256-511 byte frames transmitted"); 5775 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023", 5776 CTLFLAG_RD, &stats->ptc1023, 5777 "512-1023 byte frames transmitted"); 5778 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522", 5779 CTLFLAG_RD, &stats->ptc1522, 5780 "1024-1522 byte frames transmitted"); 5781 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tso_txd", 5782 CTLFLAG_RD, &stats->tsctc, 5783 "TSO Contexts Transmitted"); 5784 SYSCTL_ADD_QUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail", 5785 CTLFLAG_RD, &stats->tsctfc, 5786 "TSO Contexts Failed"); 5787 5788 5789 /* Interrupt Stats */ 5790 5791 int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts", 5792 CTLFLAG_RD, NULL, "Interrupt Statistics"); 5793 int_list = SYSCTL_CHILDREN(int_node); 5794 5795 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "asserts", 5796 CTLFLAG_RD, &stats->iac, 5797 "Interrupt Assertion Count"); 5798 5799 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer", 5800 CTLFLAG_RD, &stats->icrxptc, 5801 "Interrupt Cause Rx Pkt Timer Expire Count"); 5802 5803 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "rx_abs_timer", 5804 CTLFLAG_RD, &stats->icrxatc, 5805 "Interrupt Cause Rx Abs Timer Expire Count"); 5806 5807 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer", 5808 CTLFLAG_RD, &stats->ictxptc, 5809 "Interrupt Cause Tx Pkt Timer Expire Count"); 5810 5811 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "tx_abs_timer", 5812 CTLFLAG_RD, &stats->ictxatc, 5813 "Interrupt Cause Tx Abs Timer Expire Count"); 5814 5815 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "tx_queue_empty", 5816 CTLFLAG_RD, &stats->ictxqec, 5817 "Interrupt Cause Tx Queue Empty Count"); 5818 5819 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh", 5820 CTLFLAG_RD, &stats->ictxqmtc, 5821 "Interrupt Cause Tx Queue Min Thresh Count"); 5822 5823 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh", 5824 CTLFLAG_RD, &stats->icrxdmtc, 5825 "Interrupt Cause Rx Desc Min Thresh Count"); 5826 5827 SYSCTL_ADD_QUAD(ctx, int_list, OID_AUTO, "rx_overrun", 5828 CTLFLAG_RD, &stats->icrxoc, 5829 "Interrupt Cause Receiver Overrun Count"); 5830 5831 /* Host to Card Stats */ 5832 5833 host_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "host", 5834 CTLFLAG_RD, NULL, 5835 "Host to Card Statistics"); 5836 5837 host_list = SYSCTL_CHILDREN(host_node); 5838 5839 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "breaker_tx_pkt", 5840 CTLFLAG_RD, &stats->cbtmpc, 5841 "Circuit Breaker Tx Packet Count"); 5842 5843 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "host_tx_pkt_discard", 5844 CTLFLAG_RD, &stats->htdpmc, 5845 "Host Transmit Discarded Packets"); 5846 5847 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "rx_pkt", 5848 CTLFLAG_RD, &stats->rpthc, 5849 "Rx Packets To Host"); 5850 5851 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "breaker_rx_pkts", 5852 CTLFLAG_RD, &stats->cbrmpc, 5853 "Circuit Breaker Rx Packet Count"); 5854 5855 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "breaker_rx_pkt_drop", 5856 CTLFLAG_RD, &stats->cbrdpc, 5857 "Circuit Breaker Rx Dropped Count"); 5858 5859 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "tx_good_pkt", 5860 CTLFLAG_RD, &stats->hgptc, 5861 "Host Good Packets Tx Count"); 5862 5863 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "breaker_tx_pkt_drop", 5864 CTLFLAG_RD, &stats->htcbdpc, 5865 "Host Tx Circuit Breaker Dropped Count"); 5866 5867 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "rx_good_bytes", 5868 CTLFLAG_RD, &stats->hgorc, 5869 "Host Good Octets Received Count"); 5870 5871 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "tx_good_bytes", 5872 CTLFLAG_RD, &stats->hgotc, 5873 "Host Good Octets Transmit Count"); 5874 5875 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "length_errors", 5876 CTLFLAG_RD, &stats->lenerrs, 5877 "Length Errors"); 5878 5879 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "serdes_violation_pkt", 5880 CTLFLAG_RD, &stats->scvpc, 5881 "SerDes/SGMII Code Violation Pkt Count"); 5882 5883 SYSCTL_ADD_QUAD(ctx, host_list, OID_AUTO, "header_redir_missed", 5884 CTLFLAG_RD, &stats->hrmpc, 5885 "Header Redirection Missed Packet Count"); 5886} 5887 5888 5889/********************************************************************** 5890 * 5891 * This routine provides a way to dump out the adapter eeprom, 5892 * often a useful debug/service tool. This only dumps the first 5893 * 32 words, stuff that matters is in that extent. 5894 * 5895 **********************************************************************/ 5896static int 5897igb_sysctl_nvm_info(SYSCTL_HANDLER_ARGS) 5898{ 5899 struct adapter *adapter; 5900 int error; 5901 int result; 5902 5903 result = -1; 5904 error = sysctl_handle_int(oidp, &result, 0, req); 5905 5906 if (error || !req->newptr) 5907 return (error); 5908 5909 /* 5910 * This value will cause a hex dump of the 5911 * first 32 16-bit words of the EEPROM to 5912 * the screen. 5913 */ 5914 if (result == 1) { 5915 adapter = (struct adapter *)arg1; 5916 igb_print_nvm_info(adapter); 5917 } 5918 5919 return (error); 5920} 5921 5922static void 5923igb_print_nvm_info(struct adapter *adapter) 5924{ 5925 u16 eeprom_data; 5926 int i, j, row = 0; 5927 5928 /* Its a bit crude, but it gets the job done */ 5929 printf("\nInterface EEPROM Dump:\n"); 5930 printf("Offset\n0x0000 "); 5931 for (i = 0, j = 0; i < 32; i++, j++) { 5932 if (j == 8) { /* Make the offset block */ 5933 j = 0; ++row; 5934 printf("\n0x00%x0 ",row); 5935 } 5936 e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data); 5937 printf("%04x ", eeprom_data); 5938 } 5939 printf("\n"); 5940} 5941 5942static void 5943igb_set_sysctl_value(struct adapter *adapter, const char *name, 5944 const char *description, int *limit, int value) 5945{ 5946 *limit = value; 5947 SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev), 5948 SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)), 5949 OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW, limit, value, description); 5950} 5951 5952/* 5953** Set flow control using sysctl: 5954** Flow control values: 5955** 0 - off 5956** 1 - rx pause 5957** 2 - tx pause 5958** 3 - full 5959*/ 5960static int 5961igb_set_flowcntl(SYSCTL_HANDLER_ARGS) 5962{ 5963 int error; 5964 static int input = 3; /* default is full */ 5965 struct adapter *adapter = (struct adapter *) arg1; 5966 5967 error = sysctl_handle_int(oidp, &input, 0, req); 5968 5969 if ((error) || (req->newptr == NULL)) 5970 return (error); 5971 5972 switch (input) { 5973 case e1000_fc_rx_pause: 5974 case e1000_fc_tx_pause: 5975 case e1000_fc_full: 5976 case e1000_fc_none: 5977 adapter->hw.fc.requested_mode = input; 5978 adapter->fc = input; 5979 break; 5980 default: 5981 /* Do nothing */ 5982 return (error); 5983 } 5984 5985 adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode; 5986 e1000_force_mac_fc(&adapter->hw); 5987 return (error); 5988} 5989 5990/* 5991** Manage DMA Coalesce: 5992** Control values: 5993** 0/1 - off/on 5994** Legal timer values are: 5995** 250,500,1000-10000 in thousands 5996*/ 5997static int 5998igb_sysctl_dmac(SYSCTL_HANDLER_ARGS) 5999{ 6000 struct adapter *adapter = (struct adapter *) arg1; 6001 int error; 6002 6003 error = sysctl_handle_int(oidp, &adapter->dmac, 0, req); 6004 6005 if ((error) || (req->newptr == NULL)) 6006 return (error); 6007 6008 switch (adapter->dmac) { 6009 case 0: 6010 /*Disabling */ 6011 break; 6012 case 1: /* Just enable and use default */ 6013 adapter->dmac = 1000; 6014 break; 6015 case 250: 6016 case 500: 6017 case 1000: 6018 case 2000: 6019 case 3000: 6020 case 4000: 6021 case 5000: 6022 case 6000: 6023 case 7000: 6024 case 8000: 6025 case 9000: 6026 case 10000: 6027 /* Legal values - allow */ 6028 break; 6029 default: 6030 /* Do nothing, illegal value */ 6031 adapter->dmac = 0; 6032 return (error); 6033 } 6034 /* Reinit the interface */ 6035 igb_init(adapter); 6036 return (error); 6037} 6038 6039/* 6040** Manage Energy Efficient Ethernet: 6041** Control values: 6042** 0/1 - enabled/disabled 6043*/ 6044static int 6045igb_sysctl_eee(SYSCTL_HANDLER_ARGS) 6046{ 6047 struct adapter *adapter = (struct adapter *) arg1; 6048 int error, value; 6049 6050 value = adapter->hw.dev_spec._82575.eee_disable; 6051 error = sysctl_handle_int(oidp, &value, 0, req); 6052 if (error || req->newptr == NULL) 6053 return (error); 6054 IGB_CORE_LOCK(adapter); 6055 adapter->hw.dev_spec._82575.eee_disable = (value != 0); 6056 igb_init_locked(adapter); 6057 IGB_CORE_UNLOCK(adapter); 6058 return (0); 6059} 6060