if_re.c revision 217911
1/*- 2 * Copyright (c) 1997, 1998-2003 3 * Bill Paul <wpaul@windriver.com>. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by Bill Paul. 16 * 4. Neither the name of the author nor the names of any co-contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND 21 * 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 Bill Paul OR THE VOICES IN HIS HEAD 24 * BE 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 30 * THE POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33#include <sys/cdefs.h> 34__FBSDID("$FreeBSD: head/sys/dev/re/if_re.c 217911 2011-01-26 21:14:20Z yongari $"); 35 36/* 37 * RealTek 8139C+/8169/8169S/8110S/8168/8111/8101E PCI NIC driver 38 * 39 * Written by Bill Paul <wpaul@windriver.com> 40 * Senior Networking Software Engineer 41 * Wind River Systems 42 */ 43 44/* 45 * This driver is designed to support RealTek's next generation of 46 * 10/100 and 10/100/1000 PCI ethernet controllers. There are currently 47 * seven devices in this family: the RTL8139C+, the RTL8169, the RTL8169S, 48 * RTL8110S, the RTL8168, the RTL8111 and the RTL8101E. 49 * 50 * The 8139C+ is a 10/100 ethernet chip. It is backwards compatible 51 * with the older 8139 family, however it also supports a special 52 * C+ mode of operation that provides several new performance enhancing 53 * features. These include: 54 * 55 * o Descriptor based DMA mechanism. Each descriptor represents 56 * a single packet fragment. Data buffers may be aligned on 57 * any byte boundary. 58 * 59 * o 64-bit DMA 60 * 61 * o TCP/IP checksum offload for both RX and TX 62 * 63 * o High and normal priority transmit DMA rings 64 * 65 * o VLAN tag insertion and extraction 66 * 67 * o TCP large send (segmentation offload) 68 * 69 * Like the 8139, the 8139C+ also has a built-in 10/100 PHY. The C+ 70 * programming API is fairly straightforward. The RX filtering, EEPROM 71 * access and PHY access is the same as it is on the older 8139 series 72 * chips. 73 * 74 * The 8169 is a 64-bit 10/100/1000 gigabit ethernet MAC. It has almost the 75 * same programming API and feature set as the 8139C+ with the following 76 * differences and additions: 77 * 78 * o 1000Mbps mode 79 * 80 * o Jumbo frames 81 * 82 * o GMII and TBI ports/registers for interfacing with copper 83 * or fiber PHYs 84 * 85 * o RX and TX DMA rings can have up to 1024 descriptors 86 * (the 8139C+ allows a maximum of 64) 87 * 88 * o Slight differences in register layout from the 8139C+ 89 * 90 * The TX start and timer interrupt registers are at different locations 91 * on the 8169 than they are on the 8139C+. Also, the status word in the 92 * RX descriptor has a slightly different bit layout. The 8169 does not 93 * have a built-in PHY. Most reference boards use a Marvell 88E1000 'Alaska' 94 * copper gigE PHY. 95 * 96 * The 8169S/8110S 10/100/1000 devices have built-in copper gigE PHYs 97 * (the 'S' stands for 'single-chip'). These devices have the same 98 * programming API as the older 8169, but also have some vendor-specific 99 * registers for the on-board PHY. The 8110S is a LAN-on-motherboard 100 * part designed to be pin-compatible with the RealTek 8100 10/100 chip. 101 * 102 * This driver takes advantage of the RX and TX checksum offload and 103 * VLAN tag insertion/extraction features. It also implements TX 104 * interrupt moderation using the timer interrupt registers, which 105 * significantly reduces TX interrupt load. There is also support 106 * for jumbo frames, however the 8169/8169S/8110S can not transmit 107 * jumbo frames larger than 7440, so the max MTU possible with this 108 * driver is 7422 bytes. 109 */ 110 111#ifdef HAVE_KERNEL_OPTION_HEADERS 112#include "opt_device_polling.h" 113#endif 114 115#include <sys/param.h> 116#include <sys/endian.h> 117#include <sys/systm.h> 118#include <sys/sockio.h> 119#include <sys/mbuf.h> 120#include <sys/malloc.h> 121#include <sys/module.h> 122#include <sys/kernel.h> 123#include <sys/socket.h> 124#include <sys/lock.h> 125#include <sys/mutex.h> 126#include <sys/sysctl.h> 127#include <sys/taskqueue.h> 128 129#include <net/if.h> 130#include <net/if_arp.h> 131#include <net/ethernet.h> 132#include <net/if_dl.h> 133#include <net/if_media.h> 134#include <net/if_types.h> 135#include <net/if_vlan_var.h> 136 137#include <net/bpf.h> 138 139#include <machine/bus.h> 140#include <machine/resource.h> 141#include <sys/bus.h> 142#include <sys/rman.h> 143 144#include <dev/mii/mii.h> 145#include <dev/mii/miivar.h> 146 147#include <dev/pci/pcireg.h> 148#include <dev/pci/pcivar.h> 149 150#include <pci/if_rlreg.h> 151 152MODULE_DEPEND(re, pci, 1, 1, 1); 153MODULE_DEPEND(re, ether, 1, 1, 1); 154MODULE_DEPEND(re, miibus, 1, 1, 1); 155 156/* "device miibus" required. See GENERIC if you get errors here. */ 157#include "miibus_if.h" 158 159/* Tunables. */ 160static int intr_filter = 0; 161TUNABLE_INT("hw.re.intr_filter", &intr_filter); 162static int msi_disable = 0; 163TUNABLE_INT("hw.re.msi_disable", &msi_disable); 164static int msix_disable = 0; 165TUNABLE_INT("hw.re.msix_disable", &msix_disable); 166static int prefer_iomap = 0; 167TUNABLE_INT("hw.re.prefer_iomap", &prefer_iomap); 168 169#define RE_CSUM_FEATURES (CSUM_IP | CSUM_TCP | CSUM_UDP) 170 171/* 172 * Various supported device vendors/types and their names. 173 */ 174static struct rl_type re_devs[] = { 175 { DLINK_VENDORID, DLINK_DEVICEID_528T, 0, 176 "D-Link DGE-528(T) Gigabit Ethernet Adapter" }, 177 { RT_VENDORID, RT_DEVICEID_8139, 0, 178 "RealTek 8139C+ 10/100BaseTX" }, 179 { RT_VENDORID, RT_DEVICEID_8101E, 0, 180 "RealTek 810xE PCIe 10/100baseTX" }, 181 { RT_VENDORID, RT_DEVICEID_8168, 0, 182 "RealTek 8168/8111 B/C/CP/D/DP/E PCIe Gigabit Ethernet" }, 183 { RT_VENDORID, RT_DEVICEID_8169, 0, 184 "RealTek 8169/8169S/8169SB(L)/8110S/8110SB(L) Gigabit Ethernet" }, 185 { RT_VENDORID, RT_DEVICEID_8169SC, 0, 186 "RealTek 8169SC/8110SC Single-chip Gigabit Ethernet" }, 187 { COREGA_VENDORID, COREGA_DEVICEID_CGLAPCIGT, 0, 188 "Corega CG-LAPCIGT (RTL8169S) Gigabit Ethernet" }, 189 { LINKSYS_VENDORID, LINKSYS_DEVICEID_EG1032, 0, 190 "Linksys EG1032 (RTL8169S) Gigabit Ethernet" }, 191 { USR_VENDORID, USR_DEVICEID_997902, 0, 192 "US Robotics 997902 (RTL8169S) Gigabit Ethernet" } 193}; 194 195static struct rl_hwrev re_hwrevs[] = { 196 { RL_HWREV_8139, RL_8139, "", RL_MTU }, 197 { RL_HWREV_8139A, RL_8139, "A", RL_MTU }, 198 { RL_HWREV_8139AG, RL_8139, "A-G", RL_MTU }, 199 { RL_HWREV_8139B, RL_8139, "B", RL_MTU }, 200 { RL_HWREV_8130, RL_8139, "8130", RL_MTU }, 201 { RL_HWREV_8139C, RL_8139, "C", RL_MTU }, 202 { RL_HWREV_8139D, RL_8139, "8139D/8100B/8100C", RL_MTU }, 203 { RL_HWREV_8139CPLUS, RL_8139CPLUS, "C+", RL_MTU }, 204 { RL_HWREV_8168B_SPIN1, RL_8169, "8168", RL_JUMBO_MTU }, 205 { RL_HWREV_8169, RL_8169, "8169", RL_JUMBO_MTU }, 206 { RL_HWREV_8169S, RL_8169, "8169S", RL_JUMBO_MTU }, 207 { RL_HWREV_8110S, RL_8169, "8110S", RL_JUMBO_MTU }, 208 { RL_HWREV_8169_8110SB, RL_8169, "8169SB/8110SB", RL_JUMBO_MTU }, 209 { RL_HWREV_8169_8110SC, RL_8169, "8169SC/8110SC", RL_JUMBO_MTU }, 210 { RL_HWREV_8169_8110SBL, RL_8169, "8169SBL/8110SBL", RL_JUMBO_MTU }, 211 { RL_HWREV_8169_8110SCE, RL_8169, "8169SC/8110SC", RL_JUMBO_MTU }, 212 { RL_HWREV_8100, RL_8139, "8100", RL_MTU }, 213 { RL_HWREV_8101, RL_8139, "8101", RL_MTU }, 214 { RL_HWREV_8100E, RL_8169, "8100E", RL_MTU }, 215 { RL_HWREV_8101E, RL_8169, "8101E", RL_MTU }, 216 { RL_HWREV_8102E, RL_8169, "8102E", RL_MTU }, 217 { RL_HWREV_8102EL, RL_8169, "8102EL", RL_MTU }, 218 { RL_HWREV_8102EL_SPIN1, RL_8169, "8102EL", RL_MTU }, 219 { RL_HWREV_8103E, RL_8169, "8103E", RL_MTU }, 220 { RL_HWREV_8105E, RL_8169, "8105E", RL_MTU }, 221 { RL_HWREV_8168B_SPIN2, RL_8169, "8168", RL_JUMBO_MTU }, 222 { RL_HWREV_8168B_SPIN3, RL_8169, "8168", RL_JUMBO_MTU }, 223 { RL_HWREV_8168C, RL_8169, "8168C/8111C", RL_JUMBO_MTU_6K }, 224 { RL_HWREV_8168C_SPIN2, RL_8169, "8168C/8111C", RL_JUMBO_MTU_6K }, 225 { RL_HWREV_8168CP, RL_8169, "8168CP/8111CP", RL_JUMBO_MTU_6K }, 226 { RL_HWREV_8168D, RL_8169, "8168D/8111D", RL_JUMBO_MTU_9K }, 227 { RL_HWREV_8168DP, RL_8169, "8168DP/8111DP", RL_JUMBO_MTU_9K }, 228 { RL_HWREV_8168E, RL_8169, "8168E/8111E", RL_JUMBO_MTU_9K}, 229 { RL_HWREV_8168E_VL, RL_8169, "8168E/8111E-VL", RL_JUMBO_MTU_6K}, 230 { 0, 0, NULL, 0 } 231}; 232 233static int re_probe (device_t); 234static int re_attach (device_t); 235static int re_detach (device_t); 236 237static int re_encap (struct rl_softc *, struct mbuf **); 238 239static void re_dma_map_addr (void *, bus_dma_segment_t *, int, int); 240static int re_allocmem (device_t, struct rl_softc *); 241static __inline void re_discard_rxbuf 242 (struct rl_softc *, int); 243static int re_newbuf (struct rl_softc *, int); 244static int re_jumbo_newbuf (struct rl_softc *, int); 245static int re_rx_list_init (struct rl_softc *); 246static int re_jrx_list_init (struct rl_softc *); 247static int re_tx_list_init (struct rl_softc *); 248#ifdef RE_FIXUP_RX 249static __inline void re_fixup_rx 250 (struct mbuf *); 251#endif 252static int re_rxeof (struct rl_softc *, int *); 253static void re_txeof (struct rl_softc *); 254#ifdef DEVICE_POLLING 255static int re_poll (struct ifnet *, enum poll_cmd, int); 256static int re_poll_locked (struct ifnet *, enum poll_cmd, int); 257#endif 258static int re_intr (void *); 259static void re_intr_msi (void *); 260static void re_tick (void *); 261static void re_int_task (void *, int); 262static void re_start (struct ifnet *); 263static void re_start_locked (struct ifnet *); 264static int re_ioctl (struct ifnet *, u_long, caddr_t); 265static void re_init (void *); 266static void re_init_locked (struct rl_softc *); 267static void re_stop (struct rl_softc *); 268static void re_watchdog (struct rl_softc *); 269static int re_suspend (device_t); 270static int re_resume (device_t); 271static int re_shutdown (device_t); 272static int re_ifmedia_upd (struct ifnet *); 273static void re_ifmedia_sts (struct ifnet *, struct ifmediareq *); 274 275static void re_eeprom_putbyte (struct rl_softc *, int); 276static void re_eeprom_getword (struct rl_softc *, int, u_int16_t *); 277static void re_read_eeprom (struct rl_softc *, caddr_t, int, int); 278static int re_gmii_readreg (device_t, int, int); 279static int re_gmii_writereg (device_t, int, int, int); 280 281static int re_miibus_readreg (device_t, int, int); 282static int re_miibus_writereg (device_t, int, int, int); 283static void re_miibus_statchg (device_t); 284 285static void re_set_jumbo (struct rl_softc *, int); 286static void re_set_rxmode (struct rl_softc *); 287static void re_reset (struct rl_softc *); 288static void re_setwol (struct rl_softc *); 289static void re_clrwol (struct rl_softc *); 290 291#ifdef RE_DIAG 292static int re_diag (struct rl_softc *); 293#endif 294 295static void re_add_sysctls (struct rl_softc *); 296static int re_sysctl_stats (SYSCTL_HANDLER_ARGS); 297static int sysctl_int_range (SYSCTL_HANDLER_ARGS, int, int); 298static int sysctl_hw_re_int_mod (SYSCTL_HANDLER_ARGS); 299 300static device_method_t re_methods[] = { 301 /* Device interface */ 302 DEVMETHOD(device_probe, re_probe), 303 DEVMETHOD(device_attach, re_attach), 304 DEVMETHOD(device_detach, re_detach), 305 DEVMETHOD(device_suspend, re_suspend), 306 DEVMETHOD(device_resume, re_resume), 307 DEVMETHOD(device_shutdown, re_shutdown), 308 309 /* bus interface */ 310 DEVMETHOD(bus_print_child, bus_generic_print_child), 311 DEVMETHOD(bus_driver_added, bus_generic_driver_added), 312 313 /* MII interface */ 314 DEVMETHOD(miibus_readreg, re_miibus_readreg), 315 DEVMETHOD(miibus_writereg, re_miibus_writereg), 316 DEVMETHOD(miibus_statchg, re_miibus_statchg), 317 318 { 0, 0 } 319}; 320 321static driver_t re_driver = { 322 "re", 323 re_methods, 324 sizeof(struct rl_softc) 325}; 326 327static devclass_t re_devclass; 328 329DRIVER_MODULE(re, pci, re_driver, re_devclass, 0, 0); 330DRIVER_MODULE(miibus, re, miibus_driver, miibus_devclass, 0, 0); 331 332#define EE_SET(x) \ 333 CSR_WRITE_1(sc, RL_EECMD, \ 334 CSR_READ_1(sc, RL_EECMD) | x) 335 336#define EE_CLR(x) \ 337 CSR_WRITE_1(sc, RL_EECMD, \ 338 CSR_READ_1(sc, RL_EECMD) & ~x) 339 340/* 341 * Send a read command and address to the EEPROM, check for ACK. 342 */ 343static void 344re_eeprom_putbyte(struct rl_softc *sc, int addr) 345{ 346 int d, i; 347 348 d = addr | (RL_9346_READ << sc->rl_eewidth); 349 350 /* 351 * Feed in each bit and strobe the clock. 352 */ 353 354 for (i = 1 << (sc->rl_eewidth + 3); i; i >>= 1) { 355 if (d & i) { 356 EE_SET(RL_EE_DATAIN); 357 } else { 358 EE_CLR(RL_EE_DATAIN); 359 } 360 DELAY(100); 361 EE_SET(RL_EE_CLK); 362 DELAY(150); 363 EE_CLR(RL_EE_CLK); 364 DELAY(100); 365 } 366} 367 368/* 369 * Read a word of data stored in the EEPROM at address 'addr.' 370 */ 371static void 372re_eeprom_getword(struct rl_softc *sc, int addr, u_int16_t *dest) 373{ 374 int i; 375 u_int16_t word = 0; 376 377 /* 378 * Send address of word we want to read. 379 */ 380 re_eeprom_putbyte(sc, addr); 381 382 /* 383 * Start reading bits from EEPROM. 384 */ 385 for (i = 0x8000; i; i >>= 1) { 386 EE_SET(RL_EE_CLK); 387 DELAY(100); 388 if (CSR_READ_1(sc, RL_EECMD) & RL_EE_DATAOUT) 389 word |= i; 390 EE_CLR(RL_EE_CLK); 391 DELAY(100); 392 } 393 394 *dest = word; 395} 396 397/* 398 * Read a sequence of words from the EEPROM. 399 */ 400static void 401re_read_eeprom(struct rl_softc *sc, caddr_t dest, int off, int cnt) 402{ 403 int i; 404 u_int16_t word = 0, *ptr; 405 406 CSR_SETBIT_1(sc, RL_EECMD, RL_EEMODE_PROGRAM); 407 408 DELAY(100); 409 410 for (i = 0; i < cnt; i++) { 411 CSR_SETBIT_1(sc, RL_EECMD, RL_EE_SEL); 412 re_eeprom_getword(sc, off + i, &word); 413 CSR_CLRBIT_1(sc, RL_EECMD, RL_EE_SEL); 414 ptr = (u_int16_t *)(dest + (i * 2)); 415 *ptr = word; 416 } 417 418 CSR_CLRBIT_1(sc, RL_EECMD, RL_EEMODE_PROGRAM); 419} 420 421static int 422re_gmii_readreg(device_t dev, int phy, int reg) 423{ 424 struct rl_softc *sc; 425 u_int32_t rval; 426 int i; 427 428 sc = device_get_softc(dev); 429 430 /* Let the rgephy driver read the GMEDIASTAT register */ 431 432 if (reg == RL_GMEDIASTAT) { 433 rval = CSR_READ_1(sc, RL_GMEDIASTAT); 434 return (rval); 435 } 436 437 CSR_WRITE_4(sc, RL_PHYAR, reg << 16); 438 439 for (i = 0; i < RL_PHY_TIMEOUT; i++) { 440 rval = CSR_READ_4(sc, RL_PHYAR); 441 if (rval & RL_PHYAR_BUSY) 442 break; 443 DELAY(25); 444 } 445 446 if (i == RL_PHY_TIMEOUT) { 447 device_printf(sc->rl_dev, "PHY read failed\n"); 448 return (0); 449 } 450 451 /* 452 * Controller requires a 20us delay to process next MDIO request. 453 */ 454 DELAY(20); 455 456 return (rval & RL_PHYAR_PHYDATA); 457} 458 459static int 460re_gmii_writereg(device_t dev, int phy, int reg, int data) 461{ 462 struct rl_softc *sc; 463 u_int32_t rval; 464 int i; 465 466 sc = device_get_softc(dev); 467 468 CSR_WRITE_4(sc, RL_PHYAR, (reg << 16) | 469 (data & RL_PHYAR_PHYDATA) | RL_PHYAR_BUSY); 470 471 for (i = 0; i < RL_PHY_TIMEOUT; i++) { 472 rval = CSR_READ_4(sc, RL_PHYAR); 473 if (!(rval & RL_PHYAR_BUSY)) 474 break; 475 DELAY(25); 476 } 477 478 if (i == RL_PHY_TIMEOUT) { 479 device_printf(sc->rl_dev, "PHY write failed\n"); 480 return (0); 481 } 482 483 /* 484 * Controller requires a 20us delay to process next MDIO request. 485 */ 486 DELAY(20); 487 488 return (0); 489} 490 491static int 492re_miibus_readreg(device_t dev, int phy, int reg) 493{ 494 struct rl_softc *sc; 495 u_int16_t rval = 0; 496 u_int16_t re8139_reg = 0; 497 498 sc = device_get_softc(dev); 499 500 if (sc->rl_type == RL_8169) { 501 rval = re_gmii_readreg(dev, phy, reg); 502 return (rval); 503 } 504 505 switch (reg) { 506 case MII_BMCR: 507 re8139_reg = RL_BMCR; 508 break; 509 case MII_BMSR: 510 re8139_reg = RL_BMSR; 511 break; 512 case MII_ANAR: 513 re8139_reg = RL_ANAR; 514 break; 515 case MII_ANER: 516 re8139_reg = RL_ANER; 517 break; 518 case MII_ANLPAR: 519 re8139_reg = RL_LPAR; 520 break; 521 case MII_PHYIDR1: 522 case MII_PHYIDR2: 523 return (0); 524 /* 525 * Allow the rlphy driver to read the media status 526 * register. If we have a link partner which does not 527 * support NWAY, this is the register which will tell 528 * us the results of parallel detection. 529 */ 530 case RL_MEDIASTAT: 531 rval = CSR_READ_1(sc, RL_MEDIASTAT); 532 return (rval); 533 default: 534 device_printf(sc->rl_dev, "bad phy register\n"); 535 return (0); 536 } 537 rval = CSR_READ_2(sc, re8139_reg); 538 if (sc->rl_type == RL_8139CPLUS && re8139_reg == RL_BMCR) { 539 /* 8139C+ has different bit layout. */ 540 rval &= ~(BMCR_LOOP | BMCR_ISO); 541 } 542 return (rval); 543} 544 545static int 546re_miibus_writereg(device_t dev, int phy, int reg, int data) 547{ 548 struct rl_softc *sc; 549 u_int16_t re8139_reg = 0; 550 int rval = 0; 551 552 sc = device_get_softc(dev); 553 554 if (sc->rl_type == RL_8169) { 555 rval = re_gmii_writereg(dev, phy, reg, data); 556 return (rval); 557 } 558 559 switch (reg) { 560 case MII_BMCR: 561 re8139_reg = RL_BMCR; 562 if (sc->rl_type == RL_8139CPLUS) { 563 /* 8139C+ has different bit layout. */ 564 data &= ~(BMCR_LOOP | BMCR_ISO); 565 } 566 break; 567 case MII_BMSR: 568 re8139_reg = RL_BMSR; 569 break; 570 case MII_ANAR: 571 re8139_reg = RL_ANAR; 572 break; 573 case MII_ANER: 574 re8139_reg = RL_ANER; 575 break; 576 case MII_ANLPAR: 577 re8139_reg = RL_LPAR; 578 break; 579 case MII_PHYIDR1: 580 case MII_PHYIDR2: 581 return (0); 582 break; 583 default: 584 device_printf(sc->rl_dev, "bad phy register\n"); 585 return (0); 586 } 587 CSR_WRITE_2(sc, re8139_reg, data); 588 return (0); 589} 590 591static void 592re_miibus_statchg(device_t dev) 593{ 594 struct rl_softc *sc; 595 struct ifnet *ifp; 596 struct mii_data *mii; 597 598 sc = device_get_softc(dev); 599 mii = device_get_softc(sc->rl_miibus); 600 ifp = sc->rl_ifp; 601 if (mii == NULL || ifp == NULL || 602 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 603 return; 604 605 sc->rl_flags &= ~RL_FLAG_LINK; 606 if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) == 607 (IFM_ACTIVE | IFM_AVALID)) { 608 switch (IFM_SUBTYPE(mii->mii_media_active)) { 609 case IFM_10_T: 610 case IFM_100_TX: 611 sc->rl_flags |= RL_FLAG_LINK; 612 break; 613 case IFM_1000_T: 614 if ((sc->rl_flags & RL_FLAG_FASTETHER) != 0) 615 break; 616 sc->rl_flags |= RL_FLAG_LINK; 617 break; 618 default: 619 break; 620 } 621 } 622 /* 623 * RealTek controllers does not provide any interface to 624 * Tx/Rx MACs for resolved speed, duplex and flow-control 625 * parameters. 626 */ 627} 628 629/* 630 * Set the RX configuration and 64-bit multicast hash filter. 631 */ 632static void 633re_set_rxmode(struct rl_softc *sc) 634{ 635 struct ifnet *ifp; 636 struct ifmultiaddr *ifma; 637 uint32_t hashes[2] = { 0, 0 }; 638 uint32_t h, rxfilt; 639 640 RL_LOCK_ASSERT(sc); 641 642 ifp = sc->rl_ifp; 643 644 rxfilt = RL_RXCFG_CONFIG | RL_RXCFG_RX_INDIV | RL_RXCFG_RX_BROAD; 645 646 if (ifp->if_flags & (IFF_ALLMULTI | IFF_PROMISC)) { 647 if (ifp->if_flags & IFF_PROMISC) 648 rxfilt |= RL_RXCFG_RX_ALLPHYS; 649 /* 650 * Unlike other hardwares, we have to explicitly set 651 * RL_RXCFG_RX_MULTI to receive multicast frames in 652 * promiscuous mode. 653 */ 654 rxfilt |= RL_RXCFG_RX_MULTI; 655 hashes[0] = hashes[1] = 0xffffffff; 656 goto done; 657 } 658 659 if_maddr_rlock(ifp); 660 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 661 if (ifma->ifma_addr->sa_family != AF_LINK) 662 continue; 663 h = ether_crc32_be(LLADDR((struct sockaddr_dl *) 664 ifma->ifma_addr), ETHER_ADDR_LEN) >> 26; 665 if (h < 32) 666 hashes[0] |= (1 << h); 667 else 668 hashes[1] |= (1 << (h - 32)); 669 } 670 if_maddr_runlock(ifp); 671 672 if (hashes[0] != 0 || hashes[1] != 0) { 673 /* 674 * For some unfathomable reason, RealTek decided to 675 * reverse the order of the multicast hash registers 676 * in the PCI Express parts. This means we have to 677 * write the hash pattern in reverse order for those 678 * devices. 679 */ 680 if ((sc->rl_flags & RL_FLAG_PCIE) != 0) { 681 h = bswap32(hashes[0]); 682 hashes[0] = bswap32(hashes[1]); 683 hashes[1] = h; 684 } 685 rxfilt |= RL_RXCFG_RX_MULTI; 686 } 687 688done: 689 CSR_WRITE_4(sc, RL_MAR0, hashes[0]); 690 CSR_WRITE_4(sc, RL_MAR4, hashes[1]); 691 CSR_WRITE_4(sc, RL_RXCFG, rxfilt); 692} 693 694static void 695re_reset(struct rl_softc *sc) 696{ 697 int i; 698 699 RL_LOCK_ASSERT(sc); 700 701 CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_RESET); 702 703 for (i = 0; i < RL_TIMEOUT; i++) { 704 DELAY(10); 705 if (!(CSR_READ_1(sc, RL_COMMAND) & RL_CMD_RESET)) 706 break; 707 } 708 if (i == RL_TIMEOUT) 709 device_printf(sc->rl_dev, "reset never completed!\n"); 710 711 if ((sc->rl_flags & RL_FLAG_MACRESET) != 0) 712 CSR_WRITE_1(sc, 0x82, 1); 713 if (sc->rl_hwrev->rl_rev == RL_HWREV_8169S) 714 re_gmii_writereg(sc->rl_dev, 1, 0x0b, 0); 715} 716 717#ifdef RE_DIAG 718 719/* 720 * The following routine is designed to test for a defect on some 721 * 32-bit 8169 cards. Some of these NICs have the REQ64# and ACK64# 722 * lines connected to the bus, however for a 32-bit only card, they 723 * should be pulled high. The result of this defect is that the 724 * NIC will not work right if you plug it into a 64-bit slot: DMA 725 * operations will be done with 64-bit transfers, which will fail 726 * because the 64-bit data lines aren't connected. 727 * 728 * There's no way to work around this (short of talking a soldering 729 * iron to the board), however we can detect it. The method we use 730 * here is to put the NIC into digital loopback mode, set the receiver 731 * to promiscuous mode, and then try to send a frame. We then compare 732 * the frame data we sent to what was received. If the data matches, 733 * then the NIC is working correctly, otherwise we know the user has 734 * a defective NIC which has been mistakenly plugged into a 64-bit PCI 735 * slot. In the latter case, there's no way the NIC can work correctly, 736 * so we print out a message on the console and abort the device attach. 737 */ 738 739static int 740re_diag(struct rl_softc *sc) 741{ 742 struct ifnet *ifp = sc->rl_ifp; 743 struct mbuf *m0; 744 struct ether_header *eh; 745 struct rl_desc *cur_rx; 746 u_int16_t status; 747 u_int32_t rxstat; 748 int total_len, i, error = 0, phyaddr; 749 u_int8_t dst[] = { 0x00, 'h', 'e', 'l', 'l', 'o' }; 750 u_int8_t src[] = { 0x00, 'w', 'o', 'r', 'l', 'd' }; 751 752 /* Allocate a single mbuf */ 753 MGETHDR(m0, M_DONTWAIT, MT_DATA); 754 if (m0 == NULL) 755 return (ENOBUFS); 756 757 RL_LOCK(sc); 758 759 /* 760 * Initialize the NIC in test mode. This sets the chip up 761 * so that it can send and receive frames, but performs the 762 * following special functions: 763 * - Puts receiver in promiscuous mode 764 * - Enables digital loopback mode 765 * - Leaves interrupts turned off 766 */ 767 768 ifp->if_flags |= IFF_PROMISC; 769 sc->rl_testmode = 1; 770 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 771 re_init_locked(sc); 772 sc->rl_flags |= RL_FLAG_LINK; 773 if (sc->rl_type == RL_8169) 774 phyaddr = 1; 775 else 776 phyaddr = 0; 777 778 re_miibus_writereg(sc->rl_dev, phyaddr, MII_BMCR, BMCR_RESET); 779 for (i = 0; i < RL_TIMEOUT; i++) { 780 status = re_miibus_readreg(sc->rl_dev, phyaddr, MII_BMCR); 781 if (!(status & BMCR_RESET)) 782 break; 783 } 784 785 re_miibus_writereg(sc->rl_dev, phyaddr, MII_BMCR, BMCR_LOOP); 786 CSR_WRITE_2(sc, RL_ISR, RL_INTRS); 787 788 DELAY(100000); 789 790 /* Put some data in the mbuf */ 791 792 eh = mtod(m0, struct ether_header *); 793 bcopy ((char *)&dst, eh->ether_dhost, ETHER_ADDR_LEN); 794 bcopy ((char *)&src, eh->ether_shost, ETHER_ADDR_LEN); 795 eh->ether_type = htons(ETHERTYPE_IP); 796 m0->m_pkthdr.len = m0->m_len = ETHER_MIN_LEN - ETHER_CRC_LEN; 797 798 /* 799 * Queue the packet, start transmission. 800 * Note: IF_HANDOFF() ultimately calls re_start() for us. 801 */ 802 803 CSR_WRITE_2(sc, RL_ISR, 0xFFFF); 804 RL_UNLOCK(sc); 805 /* XXX: re_diag must not be called when in ALTQ mode */ 806 IF_HANDOFF(&ifp->if_snd, m0, ifp); 807 RL_LOCK(sc); 808 m0 = NULL; 809 810 /* Wait for it to propagate through the chip */ 811 812 DELAY(100000); 813 for (i = 0; i < RL_TIMEOUT; i++) { 814 status = CSR_READ_2(sc, RL_ISR); 815 CSR_WRITE_2(sc, RL_ISR, status); 816 if ((status & (RL_ISR_TIMEOUT_EXPIRED|RL_ISR_RX_OK)) == 817 (RL_ISR_TIMEOUT_EXPIRED|RL_ISR_RX_OK)) 818 break; 819 DELAY(10); 820 } 821 822 if (i == RL_TIMEOUT) { 823 device_printf(sc->rl_dev, 824 "diagnostic failed, failed to receive packet in" 825 " loopback mode\n"); 826 error = EIO; 827 goto done; 828 } 829 830 /* 831 * The packet should have been dumped into the first 832 * entry in the RX DMA ring. Grab it from there. 833 */ 834 835 bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag, 836 sc->rl_ldata.rl_rx_list_map, 837 BUS_DMASYNC_POSTREAD); 838 bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, 839 sc->rl_ldata.rl_rx_desc[0].rx_dmamap, 840 BUS_DMASYNC_POSTREAD); 841 bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag, 842 sc->rl_ldata.rl_rx_desc[0].rx_dmamap); 843 844 m0 = sc->rl_ldata.rl_rx_desc[0].rx_m; 845 sc->rl_ldata.rl_rx_desc[0].rx_m = NULL; 846 eh = mtod(m0, struct ether_header *); 847 848 cur_rx = &sc->rl_ldata.rl_rx_list[0]; 849 total_len = RL_RXBYTES(cur_rx); 850 rxstat = le32toh(cur_rx->rl_cmdstat); 851 852 if (total_len != ETHER_MIN_LEN) { 853 device_printf(sc->rl_dev, 854 "diagnostic failed, received short packet\n"); 855 error = EIO; 856 goto done; 857 } 858 859 /* Test that the received packet data matches what we sent. */ 860 861 if (bcmp((char *)&eh->ether_dhost, (char *)&dst, ETHER_ADDR_LEN) || 862 bcmp((char *)&eh->ether_shost, (char *)&src, ETHER_ADDR_LEN) || 863 ntohs(eh->ether_type) != ETHERTYPE_IP) { 864 device_printf(sc->rl_dev, "WARNING, DMA FAILURE!\n"); 865 device_printf(sc->rl_dev, "expected TX data: %6D/%6D/0x%x\n", 866 dst, ":", src, ":", ETHERTYPE_IP); 867 device_printf(sc->rl_dev, "received RX data: %6D/%6D/0x%x\n", 868 eh->ether_dhost, ":", eh->ether_shost, ":", 869 ntohs(eh->ether_type)); 870 device_printf(sc->rl_dev, "You may have a defective 32-bit " 871 "NIC plugged into a 64-bit PCI slot.\n"); 872 device_printf(sc->rl_dev, "Please re-install the NIC in a " 873 "32-bit slot for proper operation.\n"); 874 device_printf(sc->rl_dev, "Read the re(4) man page for more " 875 "details.\n"); 876 error = EIO; 877 } 878 879done: 880 /* Turn interface off, release resources */ 881 882 sc->rl_testmode = 0; 883 sc->rl_flags &= ~RL_FLAG_LINK; 884 ifp->if_flags &= ~IFF_PROMISC; 885 re_stop(sc); 886 if (m0 != NULL) 887 m_freem(m0); 888 889 RL_UNLOCK(sc); 890 891 return (error); 892} 893 894#endif 895 896/* 897 * Probe for a RealTek 8139C+/8169/8110 chip. Check the PCI vendor and device 898 * IDs against our list and return a device name if we find a match. 899 */ 900static int 901re_probe(device_t dev) 902{ 903 struct rl_type *t; 904 uint16_t devid, vendor; 905 uint16_t revid, sdevid; 906 int i; 907 908 vendor = pci_get_vendor(dev); 909 devid = pci_get_device(dev); 910 revid = pci_get_revid(dev); 911 sdevid = pci_get_subdevice(dev); 912 913 if (vendor == LINKSYS_VENDORID && devid == LINKSYS_DEVICEID_EG1032) { 914 if (sdevid != LINKSYS_SUBDEVICE_EG1032_REV3) { 915 /* 916 * Only attach to rev. 3 of the Linksys EG1032 adapter. 917 * Rev. 2 is supported by sk(4). 918 */ 919 return (ENXIO); 920 } 921 } 922 923 if (vendor == RT_VENDORID && devid == RT_DEVICEID_8139) { 924 if (revid != 0x20) { 925 /* 8139, let rl(4) take care of this device. */ 926 return (ENXIO); 927 } 928 } 929 930 t = re_devs; 931 for (i = 0; i < sizeof(re_devs) / sizeof(re_devs[0]); i++, t++) { 932 if (vendor == t->rl_vid && devid == t->rl_did) { 933 device_set_desc(dev, t->rl_name); 934 return (BUS_PROBE_DEFAULT); 935 } 936 } 937 938 return (ENXIO); 939} 940 941/* 942 * Map a single buffer address. 943 */ 944 945static void 946re_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error) 947{ 948 bus_addr_t *addr; 949 950 if (error) 951 return; 952 953 KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg)); 954 addr = arg; 955 *addr = segs->ds_addr; 956} 957 958static int 959re_allocmem(device_t dev, struct rl_softc *sc) 960{ 961 bus_addr_t lowaddr; 962 bus_size_t rx_list_size, tx_list_size; 963 int error; 964 int i; 965 966 rx_list_size = sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc); 967 tx_list_size = sc->rl_ldata.rl_tx_desc_cnt * sizeof(struct rl_desc); 968 969 /* 970 * Allocate the parent bus DMA tag appropriate for PCI. 971 * In order to use DAC, RL_CPLUSCMD_PCI_DAC bit of RL_CPLUS_CMD 972 * register should be set. However some RealTek chips are known 973 * to be buggy on DAC handling, therefore disable DAC by limiting 974 * DMA address space to 32bit. PCIe variants of RealTek chips 975 * may not have the limitation. 976 */ 977 lowaddr = BUS_SPACE_MAXADDR; 978 if ((sc->rl_flags & RL_FLAG_PCIE) == 0) 979 lowaddr = BUS_SPACE_MAXADDR_32BIT; 980 error = bus_dma_tag_create(bus_get_dma_tag(dev), 1, 0, 981 lowaddr, BUS_SPACE_MAXADDR, NULL, NULL, 982 BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0, 983 NULL, NULL, &sc->rl_parent_tag); 984 if (error) { 985 device_printf(dev, "could not allocate parent DMA tag\n"); 986 return (error); 987 } 988 989 /* 990 * Allocate map for TX mbufs. 991 */ 992 error = bus_dma_tag_create(sc->rl_parent_tag, 1, 0, 993 BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, 994 NULL, MCLBYTES * RL_NTXSEGS, RL_NTXSEGS, 4096, 0, 995 NULL, NULL, &sc->rl_ldata.rl_tx_mtag); 996 if (error) { 997 device_printf(dev, "could not allocate TX DMA tag\n"); 998 return (error); 999 } 1000 1001 /* 1002 * Allocate map for RX mbufs. 1003 */ 1004 1005 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) { 1006 error = bus_dma_tag_create(sc->rl_parent_tag, sizeof(uint64_t), 1007 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, 1008 MJUM9BYTES, 1, MJUM9BYTES, 0, NULL, NULL, 1009 &sc->rl_ldata.rl_jrx_mtag); 1010 if (error) { 1011 device_printf(dev, 1012 "could not allocate jumbo RX DMA tag\n"); 1013 return (error); 1014 } 1015 } 1016 error = bus_dma_tag_create(sc->rl_parent_tag, sizeof(uint64_t), 0, 1017 BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, 1018 MCLBYTES, 1, MCLBYTES, 0, NULL, NULL, &sc->rl_ldata.rl_rx_mtag); 1019 if (error) { 1020 device_printf(dev, "could not allocate RX DMA tag\n"); 1021 return (error); 1022 } 1023 1024 /* 1025 * Allocate map for TX descriptor list. 1026 */ 1027 error = bus_dma_tag_create(sc->rl_parent_tag, RL_RING_ALIGN, 1028 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, 1029 NULL, tx_list_size, 1, tx_list_size, 0, 1030 NULL, NULL, &sc->rl_ldata.rl_tx_list_tag); 1031 if (error) { 1032 device_printf(dev, "could not allocate TX DMA ring tag\n"); 1033 return (error); 1034 } 1035 1036 /* Allocate DMA'able memory for the TX ring */ 1037 1038 error = bus_dmamem_alloc(sc->rl_ldata.rl_tx_list_tag, 1039 (void **)&sc->rl_ldata.rl_tx_list, 1040 BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO, 1041 &sc->rl_ldata.rl_tx_list_map); 1042 if (error) { 1043 device_printf(dev, "could not allocate TX DMA ring\n"); 1044 return (error); 1045 } 1046 1047 /* Load the map for the TX ring. */ 1048 1049 sc->rl_ldata.rl_tx_list_addr = 0; 1050 error = bus_dmamap_load(sc->rl_ldata.rl_tx_list_tag, 1051 sc->rl_ldata.rl_tx_list_map, sc->rl_ldata.rl_tx_list, 1052 tx_list_size, re_dma_map_addr, 1053 &sc->rl_ldata.rl_tx_list_addr, BUS_DMA_NOWAIT); 1054 if (error != 0 || sc->rl_ldata.rl_tx_list_addr == 0) { 1055 device_printf(dev, "could not load TX DMA ring\n"); 1056 return (ENOMEM); 1057 } 1058 1059 /* Create DMA maps for TX buffers */ 1060 1061 for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) { 1062 error = bus_dmamap_create(sc->rl_ldata.rl_tx_mtag, 0, 1063 &sc->rl_ldata.rl_tx_desc[i].tx_dmamap); 1064 if (error) { 1065 device_printf(dev, "could not create DMA map for TX\n"); 1066 return (error); 1067 } 1068 } 1069 1070 /* 1071 * Allocate map for RX descriptor list. 1072 */ 1073 error = bus_dma_tag_create(sc->rl_parent_tag, RL_RING_ALIGN, 1074 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, 1075 NULL, rx_list_size, 1, rx_list_size, 0, 1076 NULL, NULL, &sc->rl_ldata.rl_rx_list_tag); 1077 if (error) { 1078 device_printf(dev, "could not create RX DMA ring tag\n"); 1079 return (error); 1080 } 1081 1082 /* Allocate DMA'able memory for the RX ring */ 1083 1084 error = bus_dmamem_alloc(sc->rl_ldata.rl_rx_list_tag, 1085 (void **)&sc->rl_ldata.rl_rx_list, 1086 BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO, 1087 &sc->rl_ldata.rl_rx_list_map); 1088 if (error) { 1089 device_printf(dev, "could not allocate RX DMA ring\n"); 1090 return (error); 1091 } 1092 1093 /* Load the map for the RX ring. */ 1094 1095 sc->rl_ldata.rl_rx_list_addr = 0; 1096 error = bus_dmamap_load(sc->rl_ldata.rl_rx_list_tag, 1097 sc->rl_ldata.rl_rx_list_map, sc->rl_ldata.rl_rx_list, 1098 rx_list_size, re_dma_map_addr, 1099 &sc->rl_ldata.rl_rx_list_addr, BUS_DMA_NOWAIT); 1100 if (error != 0 || sc->rl_ldata.rl_rx_list_addr == 0) { 1101 device_printf(dev, "could not load RX DMA ring\n"); 1102 return (ENOMEM); 1103 } 1104 1105 /* Create DMA maps for RX buffers */ 1106 1107 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) { 1108 error = bus_dmamap_create(sc->rl_ldata.rl_jrx_mtag, 0, 1109 &sc->rl_ldata.rl_jrx_sparemap); 1110 if (error) { 1111 device_printf(dev, 1112 "could not create spare DMA map for jumbo RX\n"); 1113 return (error); 1114 } 1115 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 1116 error = bus_dmamap_create(sc->rl_ldata.rl_jrx_mtag, 0, 1117 &sc->rl_ldata.rl_jrx_desc[i].rx_dmamap); 1118 if (error) { 1119 device_printf(dev, 1120 "could not create DMA map for jumbo RX\n"); 1121 return (error); 1122 } 1123 } 1124 } 1125 error = bus_dmamap_create(sc->rl_ldata.rl_rx_mtag, 0, 1126 &sc->rl_ldata.rl_rx_sparemap); 1127 if (error) { 1128 device_printf(dev, "could not create spare DMA map for RX\n"); 1129 return (error); 1130 } 1131 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 1132 error = bus_dmamap_create(sc->rl_ldata.rl_rx_mtag, 0, 1133 &sc->rl_ldata.rl_rx_desc[i].rx_dmamap); 1134 if (error) { 1135 device_printf(dev, "could not create DMA map for RX\n"); 1136 return (error); 1137 } 1138 } 1139 1140 /* Create DMA map for statistics. */ 1141 error = bus_dma_tag_create(sc->rl_parent_tag, RL_DUMP_ALIGN, 0, 1142 BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, 1143 sizeof(struct rl_stats), 1, sizeof(struct rl_stats), 0, NULL, NULL, 1144 &sc->rl_ldata.rl_stag); 1145 if (error) { 1146 device_printf(dev, "could not create statistics DMA tag\n"); 1147 return (error); 1148 } 1149 /* Allocate DMA'able memory for statistics. */ 1150 error = bus_dmamem_alloc(sc->rl_ldata.rl_stag, 1151 (void **)&sc->rl_ldata.rl_stats, 1152 BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO, 1153 &sc->rl_ldata.rl_smap); 1154 if (error) { 1155 device_printf(dev, 1156 "could not allocate statistics DMA memory\n"); 1157 return (error); 1158 } 1159 /* Load the map for statistics. */ 1160 sc->rl_ldata.rl_stats_addr = 0; 1161 error = bus_dmamap_load(sc->rl_ldata.rl_stag, sc->rl_ldata.rl_smap, 1162 sc->rl_ldata.rl_stats, sizeof(struct rl_stats), re_dma_map_addr, 1163 &sc->rl_ldata.rl_stats_addr, BUS_DMA_NOWAIT); 1164 if (error != 0 || sc->rl_ldata.rl_stats_addr == 0) { 1165 device_printf(dev, "could not load statistics DMA memory\n"); 1166 return (ENOMEM); 1167 } 1168 1169 return (0); 1170} 1171 1172/* 1173 * Attach the interface. Allocate softc structures, do ifmedia 1174 * setup and ethernet/BPF attach. 1175 */ 1176static int 1177re_attach(device_t dev) 1178{ 1179 u_char eaddr[ETHER_ADDR_LEN]; 1180 u_int16_t as[ETHER_ADDR_LEN / 2]; 1181 struct rl_softc *sc; 1182 struct ifnet *ifp; 1183 struct rl_hwrev *hw_rev; 1184 int hwrev; 1185 u_int16_t devid, re_did = 0; 1186 int error = 0, i, phy, rid; 1187 int msic, msixc, reg; 1188 uint8_t cfg; 1189 1190 sc = device_get_softc(dev); 1191 sc->rl_dev = dev; 1192 1193 mtx_init(&sc->rl_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK, 1194 MTX_DEF); 1195 callout_init_mtx(&sc->rl_stat_callout, &sc->rl_mtx, 0); 1196 1197 /* 1198 * Map control/status registers. 1199 */ 1200 pci_enable_busmaster(dev); 1201 1202 devid = pci_get_device(dev); 1203 /* 1204 * Prefer memory space register mapping over IO space. 1205 * Because RTL8169SC does not seem to work when memory mapping 1206 * is used always activate io mapping. 1207 */ 1208 if (devid == RT_DEVICEID_8169SC) 1209 prefer_iomap = 1; 1210 if (prefer_iomap == 0) { 1211 sc->rl_res_id = PCIR_BAR(1); 1212 sc->rl_res_type = SYS_RES_MEMORY; 1213 /* RTL8168/8101E seems to use different BARs. */ 1214 if (devid == RT_DEVICEID_8168 || devid == RT_DEVICEID_8101E) 1215 sc->rl_res_id = PCIR_BAR(2); 1216 } else { 1217 sc->rl_res_id = PCIR_BAR(0); 1218 sc->rl_res_type = SYS_RES_IOPORT; 1219 } 1220 sc->rl_res = bus_alloc_resource_any(dev, sc->rl_res_type, 1221 &sc->rl_res_id, RF_ACTIVE); 1222 if (sc->rl_res == NULL && prefer_iomap == 0) { 1223 sc->rl_res_id = PCIR_BAR(0); 1224 sc->rl_res_type = SYS_RES_IOPORT; 1225 sc->rl_res = bus_alloc_resource_any(dev, sc->rl_res_type, 1226 &sc->rl_res_id, RF_ACTIVE); 1227 } 1228 if (sc->rl_res == NULL) { 1229 device_printf(dev, "couldn't map ports/memory\n"); 1230 error = ENXIO; 1231 goto fail; 1232 } 1233 1234 sc->rl_btag = rman_get_bustag(sc->rl_res); 1235 sc->rl_bhandle = rman_get_bushandle(sc->rl_res); 1236 1237 msic = pci_msi_count(dev); 1238 msixc = pci_msix_count(dev); 1239 if (pci_find_extcap(dev, PCIY_EXPRESS, ®) == 0) 1240 sc->rl_flags |= RL_FLAG_PCIE; 1241 if (bootverbose) { 1242 device_printf(dev, "MSI count : %d\n", msic); 1243 device_printf(dev, "MSI-X count : %d\n", msixc); 1244 } 1245 if (msix_disable > 0) 1246 msixc = 0; 1247 if (msi_disable > 0) 1248 msic = 0; 1249 /* Prefer MSI-X to MSI. */ 1250 if (msixc > 0) { 1251 msixc = 1; 1252 rid = PCIR_BAR(4); 1253 sc->rl_res_pba = bus_alloc_resource_any(dev, SYS_RES_MEMORY, 1254 &rid, RF_ACTIVE); 1255 if (sc->rl_res_pba == NULL) { 1256 device_printf(sc->rl_dev, 1257 "could not allocate MSI-X PBA resource\n"); 1258 } 1259 if (sc->rl_res_pba != NULL && 1260 pci_alloc_msix(dev, &msixc) == 0) { 1261 if (msixc == 1) { 1262 device_printf(dev, "Using %d MSI-X message\n", 1263 msixc); 1264 sc->rl_flags |= RL_FLAG_MSIX; 1265 } else 1266 pci_release_msi(dev); 1267 } 1268 if ((sc->rl_flags & RL_FLAG_MSIX) == 0) { 1269 if (sc->rl_res_pba != NULL) 1270 bus_release_resource(dev, SYS_RES_MEMORY, rid, 1271 sc->rl_res_pba); 1272 sc->rl_res_pba = NULL; 1273 msixc = 0; 1274 } 1275 } 1276 /* Prefer MSI to INTx. */ 1277 if (msixc == 0 && msic > 0) { 1278 msic = 1; 1279 if (pci_alloc_msi(dev, &msic) == 0) { 1280 if (msic == RL_MSI_MESSAGES) { 1281 device_printf(dev, "Using %d MSI message\n", 1282 msic); 1283 sc->rl_flags |= RL_FLAG_MSI; 1284 /* Explicitly set MSI enable bit. */ 1285 CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE); 1286 cfg = CSR_READ_1(sc, RL_CFG2); 1287 cfg |= RL_CFG2_MSI; 1288 CSR_WRITE_1(sc, RL_CFG2, cfg); 1289 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 1290 } else 1291 pci_release_msi(dev); 1292 } 1293 if ((sc->rl_flags & RL_FLAG_MSI) == 0) 1294 msic = 0; 1295 } 1296 1297 /* Allocate interrupt */ 1298 if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) == 0) { 1299 rid = 0; 1300 sc->rl_irq[0] = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid, 1301 RF_SHAREABLE | RF_ACTIVE); 1302 if (sc->rl_irq[0] == NULL) { 1303 device_printf(dev, "couldn't allocate IRQ resources\n"); 1304 error = ENXIO; 1305 goto fail; 1306 } 1307 } else { 1308 for (i = 0, rid = 1; i < RL_MSI_MESSAGES; i++, rid++) { 1309 sc->rl_irq[i] = bus_alloc_resource_any(dev, 1310 SYS_RES_IRQ, &rid, RF_ACTIVE); 1311 if (sc->rl_irq[i] == NULL) { 1312 device_printf(dev, 1313 "couldn't llocate IRQ resources for " 1314 "message %d\n", rid); 1315 error = ENXIO; 1316 goto fail; 1317 } 1318 } 1319 } 1320 1321 if ((sc->rl_flags & RL_FLAG_MSI) == 0) { 1322 CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE); 1323 cfg = CSR_READ_1(sc, RL_CFG2); 1324 if ((cfg & RL_CFG2_MSI) != 0) { 1325 device_printf(dev, "turning off MSI enable bit.\n"); 1326 cfg &= ~RL_CFG2_MSI; 1327 CSR_WRITE_1(sc, RL_CFG2, cfg); 1328 } 1329 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 1330 } 1331 1332 hw_rev = re_hwrevs; 1333 hwrev = CSR_READ_4(sc, RL_TXCFG); 1334 switch (hwrev & 0x70000000) { 1335 case 0x00000000: 1336 case 0x10000000: 1337 device_printf(dev, "Chip rev. 0x%08x\n", hwrev & 0xfc800000); 1338 hwrev &= (RL_TXCFG_HWREV | 0x80000000); 1339 break; 1340 default: 1341 device_printf(dev, "Chip rev. 0x%08x\n", hwrev & 0x7c800000); 1342 hwrev &= RL_TXCFG_HWREV; 1343 break; 1344 } 1345 device_printf(dev, "MAC rev. 0x%08x\n", hwrev & 0x00700000); 1346 while (hw_rev->rl_desc != NULL) { 1347 if (hw_rev->rl_rev == hwrev) { 1348 sc->rl_type = hw_rev->rl_type; 1349 sc->rl_hwrev = hw_rev; 1350 break; 1351 } 1352 hw_rev++; 1353 } 1354 if (hw_rev->rl_desc == NULL) { 1355 device_printf(dev, "Unknown H/W revision: 0x%08x\n", hwrev); 1356 error = ENXIO; 1357 goto fail; 1358 } 1359 1360 switch (hw_rev->rl_rev) { 1361 case RL_HWREV_8139CPLUS: 1362 sc->rl_flags |= RL_FLAG_FASTETHER | RL_FLAG_AUTOPAD; 1363 break; 1364 case RL_HWREV_8100E: 1365 case RL_HWREV_8101E: 1366 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_FASTETHER; 1367 break; 1368 case RL_HWREV_8102E: 1369 case RL_HWREV_8102EL: 1370 case RL_HWREV_8102EL_SPIN1: 1371 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR | RL_FLAG_DESCV2 | 1372 RL_FLAG_MACSTAT | RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP | 1373 RL_FLAG_AUTOPAD; 1374 break; 1375 case RL_HWREV_8103E: 1376 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR | RL_FLAG_DESCV2 | 1377 RL_FLAG_MACSTAT | RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP | 1378 RL_FLAG_AUTOPAD | RL_FLAG_MACSLEEP; 1379 break; 1380 case RL_HWREV_8105E: 1381 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PHYWAKE_PM | 1382 RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | 1383 RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP | RL_FLAG_AUTOPAD; 1384 break; 1385 case RL_HWREV_8168B_SPIN1: 1386 case RL_HWREV_8168B_SPIN2: 1387 sc->rl_flags |= RL_FLAG_WOLRXENB; 1388 /* FALLTHROUGH */ 1389 case RL_HWREV_8168B_SPIN3: 1390 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_MACSTAT; 1391 break; 1392 case RL_HWREV_8168C_SPIN2: 1393 sc->rl_flags |= RL_FLAG_MACSLEEP; 1394 /* FALLTHROUGH */ 1395 case RL_HWREV_8168C: 1396 if ((hwrev & 0x00700000) == 0x00200000) 1397 sc->rl_flags |= RL_FLAG_MACSLEEP; 1398 /* FALLTHROUGH */ 1399 case RL_HWREV_8168CP: 1400 case RL_HWREV_8168D: 1401 case RL_HWREV_8168DP: 1402 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR | 1403 RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP | 1404 RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2; 1405 break; 1406 case RL_HWREV_8168E: 1407 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PHYWAKE_PM | 1408 RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | 1409 RL_FLAG_CMDSTOP | RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2; 1410 break; 1411 case RL_HWREV_8168E_VL: 1412 sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR | 1413 RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP | 1414 RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2; 1415 break; 1416 case RL_HWREV_8169_8110SB: 1417 case RL_HWREV_8169_8110SBL: 1418 case RL_HWREV_8169_8110SC: 1419 case RL_HWREV_8169_8110SCE: 1420 sc->rl_flags |= RL_FLAG_PHYWAKE; 1421 /* FALLTHROUGH */ 1422 case RL_HWREV_8169: 1423 case RL_HWREV_8169S: 1424 case RL_HWREV_8110S: 1425 sc->rl_flags |= RL_FLAG_MACRESET; 1426 break; 1427 default: 1428 break; 1429 } 1430 1431 /* Reset the adapter. */ 1432 RL_LOCK(sc); 1433 re_reset(sc); 1434 RL_UNLOCK(sc); 1435 1436 /* Enable PME. */ 1437 CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE); 1438 cfg = CSR_READ_1(sc, RL_CFG1); 1439 cfg |= RL_CFG1_PME; 1440 CSR_WRITE_1(sc, RL_CFG1, cfg); 1441 cfg = CSR_READ_1(sc, RL_CFG5); 1442 cfg &= RL_CFG5_PME_STS; 1443 CSR_WRITE_1(sc, RL_CFG5, cfg); 1444 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 1445 1446 if ((sc->rl_flags & RL_FLAG_PAR) != 0) { 1447 /* 1448 * XXX Should have a better way to extract station 1449 * address from EEPROM. 1450 */ 1451 for (i = 0; i < ETHER_ADDR_LEN; i++) 1452 eaddr[i] = CSR_READ_1(sc, RL_IDR0 + i); 1453 } else { 1454 sc->rl_eewidth = RL_9356_ADDR_LEN; 1455 re_read_eeprom(sc, (caddr_t)&re_did, 0, 1); 1456 if (re_did != 0x8129) 1457 sc->rl_eewidth = RL_9346_ADDR_LEN; 1458 1459 /* 1460 * Get station address from the EEPROM. 1461 */ 1462 re_read_eeprom(sc, (caddr_t)as, RL_EE_EADDR, 3); 1463 for (i = 0; i < ETHER_ADDR_LEN / 2; i++) 1464 as[i] = le16toh(as[i]); 1465 bcopy(as, eaddr, sizeof(eaddr)); 1466 } 1467 1468 if (sc->rl_type == RL_8169) { 1469 /* Set RX length mask and number of descriptors. */ 1470 sc->rl_rxlenmask = RL_RDESC_STAT_GFRAGLEN; 1471 sc->rl_txstart = RL_GTXSTART; 1472 sc->rl_ldata.rl_tx_desc_cnt = RL_8169_TX_DESC_CNT; 1473 sc->rl_ldata.rl_rx_desc_cnt = RL_8169_RX_DESC_CNT; 1474 } else { 1475 /* Set RX length mask and number of descriptors. */ 1476 sc->rl_rxlenmask = RL_RDESC_STAT_FRAGLEN; 1477 sc->rl_txstart = RL_TXSTART; 1478 sc->rl_ldata.rl_tx_desc_cnt = RL_8139_TX_DESC_CNT; 1479 sc->rl_ldata.rl_rx_desc_cnt = RL_8139_RX_DESC_CNT; 1480 } 1481 1482 error = re_allocmem(dev, sc); 1483 if (error) 1484 goto fail; 1485 re_add_sysctls(sc); 1486 1487 ifp = sc->rl_ifp = if_alloc(IFT_ETHER); 1488 if (ifp == NULL) { 1489 device_printf(dev, "can not if_alloc()\n"); 1490 error = ENOSPC; 1491 goto fail; 1492 } 1493 1494 /* Take controller out of deep sleep mode. */ 1495 if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) { 1496 if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80) 1497 CSR_WRITE_1(sc, RL_GPIO, 1498 CSR_READ_1(sc, RL_GPIO) | 0x01); 1499 else 1500 CSR_WRITE_1(sc, RL_GPIO, 1501 CSR_READ_1(sc, RL_GPIO) & ~0x01); 1502 } 1503 1504 /* Take PHY out of power down mode. */ 1505 if ((sc->rl_flags & RL_FLAG_PHYWAKE_PM) != 0) 1506 CSR_WRITE_1(sc, RL_PMCH, CSR_READ_1(sc, RL_PMCH) | 0x80); 1507 if ((sc->rl_flags & RL_FLAG_PHYWAKE) != 0) { 1508 re_gmii_writereg(dev, 1, 0x1f, 0); 1509 re_gmii_writereg(dev, 1, 0x0e, 0); 1510 } 1511 1512#define RE_PHYAD_INTERNAL 0 1513 1514 /* Do MII setup. */ 1515 phy = RE_PHYAD_INTERNAL; 1516 if (sc->rl_type == RL_8169) 1517 phy = 1; 1518 error = mii_attach(dev, &sc->rl_miibus, ifp, re_ifmedia_upd, 1519 re_ifmedia_sts, BMSR_DEFCAPMASK, phy, MII_OFFSET_ANY, MIIF_DOPAUSE); 1520 if (error != 0) { 1521 device_printf(dev, "attaching PHYs failed\n"); 1522 goto fail; 1523 } 1524 1525 ifp->if_softc = sc; 1526 if_initname(ifp, device_get_name(dev), device_get_unit(dev)); 1527 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 1528 ifp->if_ioctl = re_ioctl; 1529 ifp->if_start = re_start; 1530 ifp->if_hwassist = RE_CSUM_FEATURES | CSUM_TSO; 1531 ifp->if_capabilities = IFCAP_HWCSUM | IFCAP_TSO4; 1532 ifp->if_capenable = ifp->if_capabilities; 1533 ifp->if_init = re_init; 1534 IFQ_SET_MAXLEN(&ifp->if_snd, RL_IFQ_MAXLEN); 1535 ifp->if_snd.ifq_drv_maxlen = RL_IFQ_MAXLEN; 1536 IFQ_SET_READY(&ifp->if_snd); 1537 1538 TASK_INIT(&sc->rl_inttask, 0, re_int_task, sc); 1539 1540 /* 1541 * Call MI attach routine. 1542 */ 1543 ether_ifattach(ifp, eaddr); 1544 1545 /* VLAN capability setup */ 1546 ifp->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING; 1547 if (ifp->if_capabilities & IFCAP_HWCSUM) 1548 ifp->if_capabilities |= IFCAP_VLAN_HWCSUM; 1549 /* Enable WOL if PM is supported. */ 1550 if (pci_find_extcap(sc->rl_dev, PCIY_PMG, ®) == 0) 1551 ifp->if_capabilities |= IFCAP_WOL; 1552 ifp->if_capenable = ifp->if_capabilities; 1553 /* 1554 * Don't enable TSO by default. It is known to generate 1555 * corrupted TCP segments(bad TCP options) under certain 1556 * circumtances. 1557 */ 1558 ifp->if_hwassist &= ~CSUM_TSO; 1559 ifp->if_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO); 1560#ifdef DEVICE_POLLING 1561 ifp->if_capabilities |= IFCAP_POLLING; 1562#endif 1563 /* 1564 * Tell the upper layer(s) we support long frames. 1565 * Must appear after the call to ether_ifattach() because 1566 * ether_ifattach() sets ifi_hdrlen to the default value. 1567 */ 1568 ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header); 1569 1570#ifdef RE_DIAG 1571 /* 1572 * Perform hardware diagnostic on the original RTL8169. 1573 * Some 32-bit cards were incorrectly wired and would 1574 * malfunction if plugged into a 64-bit slot. 1575 */ 1576 1577 if (hwrev == RL_HWREV_8169) { 1578 error = re_diag(sc); 1579 if (error) { 1580 device_printf(dev, 1581 "attach aborted due to hardware diag failure\n"); 1582 ether_ifdetach(ifp); 1583 goto fail; 1584 } 1585 } 1586#endif 1587 1588#ifdef RE_TX_MODERATION 1589 intr_filter = 1; 1590#endif 1591 /* Hook interrupt last to avoid having to lock softc */ 1592 if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) != 0 && 1593 intr_filter == 0) { 1594 error = bus_setup_intr(dev, sc->rl_irq[0], 1595 INTR_TYPE_NET | INTR_MPSAFE, NULL, re_intr_msi, sc, 1596 &sc->rl_intrhand[0]); 1597 } else { 1598 error = bus_setup_intr(dev, sc->rl_irq[0], 1599 INTR_TYPE_NET | INTR_MPSAFE, re_intr, NULL, sc, 1600 &sc->rl_intrhand[0]); 1601 } 1602 if (error) { 1603 device_printf(dev, "couldn't set up irq\n"); 1604 ether_ifdetach(ifp); 1605 } 1606 1607fail: 1608 1609 if (error) 1610 re_detach(dev); 1611 1612 return (error); 1613} 1614 1615/* 1616 * Shutdown hardware and free up resources. This can be called any 1617 * time after the mutex has been initialized. It is called in both 1618 * the error case in attach and the normal detach case so it needs 1619 * to be careful about only freeing resources that have actually been 1620 * allocated. 1621 */ 1622static int 1623re_detach(device_t dev) 1624{ 1625 struct rl_softc *sc; 1626 struct ifnet *ifp; 1627 int i, rid; 1628 1629 sc = device_get_softc(dev); 1630 ifp = sc->rl_ifp; 1631 KASSERT(mtx_initialized(&sc->rl_mtx), ("re mutex not initialized")); 1632 1633 /* These should only be active if attach succeeded */ 1634 if (device_is_attached(dev)) { 1635#ifdef DEVICE_POLLING 1636 if (ifp->if_capenable & IFCAP_POLLING) 1637 ether_poll_deregister(ifp); 1638#endif 1639 RL_LOCK(sc); 1640#if 0 1641 sc->suspended = 1; 1642#endif 1643 re_stop(sc); 1644 RL_UNLOCK(sc); 1645 callout_drain(&sc->rl_stat_callout); 1646 taskqueue_drain(taskqueue_fast, &sc->rl_inttask); 1647 /* 1648 * Force off the IFF_UP flag here, in case someone 1649 * still had a BPF descriptor attached to this 1650 * interface. If they do, ether_ifdetach() will cause 1651 * the BPF code to try and clear the promisc mode 1652 * flag, which will bubble down to re_ioctl(), 1653 * which will try to call re_init() again. This will 1654 * turn the NIC back on and restart the MII ticker, 1655 * which will panic the system when the kernel tries 1656 * to invoke the re_tick() function that isn't there 1657 * anymore. 1658 */ 1659 ifp->if_flags &= ~IFF_UP; 1660 ether_ifdetach(ifp); 1661 } 1662 if (sc->rl_miibus) 1663 device_delete_child(dev, sc->rl_miibus); 1664 bus_generic_detach(dev); 1665 1666 /* 1667 * The rest is resource deallocation, so we should already be 1668 * stopped here. 1669 */ 1670 1671 if (sc->rl_intrhand[0] != NULL) { 1672 bus_teardown_intr(dev, sc->rl_irq[0], sc->rl_intrhand[0]); 1673 sc->rl_intrhand[0] = NULL; 1674 } 1675 if (ifp != NULL) 1676 if_free(ifp); 1677 if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) == 0) 1678 rid = 0; 1679 else 1680 rid = 1; 1681 if (sc->rl_irq[0] != NULL) { 1682 bus_release_resource(dev, SYS_RES_IRQ, rid, sc->rl_irq[0]); 1683 sc->rl_irq[0] = NULL; 1684 } 1685 if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) != 0) 1686 pci_release_msi(dev); 1687 if (sc->rl_res_pba) { 1688 rid = PCIR_BAR(4); 1689 bus_release_resource(dev, SYS_RES_MEMORY, rid, sc->rl_res_pba); 1690 } 1691 if (sc->rl_res) 1692 bus_release_resource(dev, sc->rl_res_type, sc->rl_res_id, 1693 sc->rl_res); 1694 1695 /* Unload and free the RX DMA ring memory and map */ 1696 1697 if (sc->rl_ldata.rl_rx_list_tag) { 1698 if (sc->rl_ldata.rl_rx_list_map) 1699 bus_dmamap_unload(sc->rl_ldata.rl_rx_list_tag, 1700 sc->rl_ldata.rl_rx_list_map); 1701 if (sc->rl_ldata.rl_rx_list_map && sc->rl_ldata.rl_rx_list) 1702 bus_dmamem_free(sc->rl_ldata.rl_rx_list_tag, 1703 sc->rl_ldata.rl_rx_list, 1704 sc->rl_ldata.rl_rx_list_map); 1705 bus_dma_tag_destroy(sc->rl_ldata.rl_rx_list_tag); 1706 } 1707 1708 /* Unload and free the TX DMA ring memory and map */ 1709 1710 if (sc->rl_ldata.rl_tx_list_tag) { 1711 if (sc->rl_ldata.rl_tx_list_map) 1712 bus_dmamap_unload(sc->rl_ldata.rl_tx_list_tag, 1713 sc->rl_ldata.rl_tx_list_map); 1714 if (sc->rl_ldata.rl_tx_list_map && sc->rl_ldata.rl_tx_list) 1715 bus_dmamem_free(sc->rl_ldata.rl_tx_list_tag, 1716 sc->rl_ldata.rl_tx_list, 1717 sc->rl_ldata.rl_tx_list_map); 1718 bus_dma_tag_destroy(sc->rl_ldata.rl_tx_list_tag); 1719 } 1720 1721 /* Destroy all the RX and TX buffer maps */ 1722 1723 if (sc->rl_ldata.rl_tx_mtag) { 1724 for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) { 1725 if (sc->rl_ldata.rl_tx_desc[i].tx_dmamap) 1726 bus_dmamap_destroy(sc->rl_ldata.rl_tx_mtag, 1727 sc->rl_ldata.rl_tx_desc[i].tx_dmamap); 1728 } 1729 bus_dma_tag_destroy(sc->rl_ldata.rl_tx_mtag); 1730 } 1731 if (sc->rl_ldata.rl_rx_mtag) { 1732 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 1733 if (sc->rl_ldata.rl_rx_desc[i].rx_dmamap) 1734 bus_dmamap_destroy(sc->rl_ldata.rl_rx_mtag, 1735 sc->rl_ldata.rl_rx_desc[i].rx_dmamap); 1736 } 1737 if (sc->rl_ldata.rl_rx_sparemap) 1738 bus_dmamap_destroy(sc->rl_ldata.rl_rx_mtag, 1739 sc->rl_ldata.rl_rx_sparemap); 1740 bus_dma_tag_destroy(sc->rl_ldata.rl_rx_mtag); 1741 } 1742 if (sc->rl_ldata.rl_jrx_mtag) { 1743 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 1744 if (sc->rl_ldata.rl_jrx_desc[i].rx_dmamap) 1745 bus_dmamap_destroy(sc->rl_ldata.rl_jrx_mtag, 1746 sc->rl_ldata.rl_jrx_desc[i].rx_dmamap); 1747 } 1748 if (sc->rl_ldata.rl_jrx_sparemap) 1749 bus_dmamap_destroy(sc->rl_ldata.rl_jrx_mtag, 1750 sc->rl_ldata.rl_jrx_sparemap); 1751 bus_dma_tag_destroy(sc->rl_ldata.rl_jrx_mtag); 1752 } 1753 /* Unload and free the stats buffer and map */ 1754 1755 if (sc->rl_ldata.rl_stag) { 1756 if (sc->rl_ldata.rl_smap) 1757 bus_dmamap_unload(sc->rl_ldata.rl_stag, 1758 sc->rl_ldata.rl_smap); 1759 if (sc->rl_ldata.rl_smap && sc->rl_ldata.rl_stats) 1760 bus_dmamem_free(sc->rl_ldata.rl_stag, 1761 sc->rl_ldata.rl_stats, sc->rl_ldata.rl_smap); 1762 bus_dma_tag_destroy(sc->rl_ldata.rl_stag); 1763 } 1764 1765 if (sc->rl_parent_tag) 1766 bus_dma_tag_destroy(sc->rl_parent_tag); 1767 1768 mtx_destroy(&sc->rl_mtx); 1769 1770 return (0); 1771} 1772 1773static __inline void 1774re_discard_rxbuf(struct rl_softc *sc, int idx) 1775{ 1776 struct rl_desc *desc; 1777 struct rl_rxdesc *rxd; 1778 uint32_t cmdstat; 1779 1780 if (sc->rl_ifp->if_mtu > RL_MTU && 1781 (sc->rl_flags & RL_FLAG_JUMBOV2) != 0) 1782 rxd = &sc->rl_ldata.rl_jrx_desc[idx]; 1783 else 1784 rxd = &sc->rl_ldata.rl_rx_desc[idx]; 1785 desc = &sc->rl_ldata.rl_rx_list[idx]; 1786 desc->rl_vlanctl = 0; 1787 cmdstat = rxd->rx_size; 1788 if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1) 1789 cmdstat |= RL_RDESC_CMD_EOR; 1790 desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN); 1791} 1792 1793static int 1794re_newbuf(struct rl_softc *sc, int idx) 1795{ 1796 struct mbuf *m; 1797 struct rl_rxdesc *rxd; 1798 bus_dma_segment_t segs[1]; 1799 bus_dmamap_t map; 1800 struct rl_desc *desc; 1801 uint32_t cmdstat; 1802 int error, nsegs; 1803 1804 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR); 1805 if (m == NULL) 1806 return (ENOBUFS); 1807 1808 m->m_len = m->m_pkthdr.len = MCLBYTES; 1809#ifdef RE_FIXUP_RX 1810 /* 1811 * This is part of an evil trick to deal with non-x86 platforms. 1812 * The RealTek chip requires RX buffers to be aligned on 64-bit 1813 * boundaries, but that will hose non-x86 machines. To get around 1814 * this, we leave some empty space at the start of each buffer 1815 * and for non-x86 hosts, we copy the buffer back six bytes 1816 * to achieve word alignment. This is slightly more efficient 1817 * than allocating a new buffer, copying the contents, and 1818 * discarding the old buffer. 1819 */ 1820 m_adj(m, RE_ETHER_ALIGN); 1821#endif 1822 error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_rx_mtag, 1823 sc->rl_ldata.rl_rx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT); 1824 if (error != 0) { 1825 m_freem(m); 1826 return (ENOBUFS); 1827 } 1828 KASSERT(nsegs == 1, ("%s: %d segment returned!", __func__, nsegs)); 1829 1830 rxd = &sc->rl_ldata.rl_rx_desc[idx]; 1831 if (rxd->rx_m != NULL) { 1832 bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap, 1833 BUS_DMASYNC_POSTREAD); 1834 bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap); 1835 } 1836 1837 rxd->rx_m = m; 1838 map = rxd->rx_dmamap; 1839 rxd->rx_dmamap = sc->rl_ldata.rl_rx_sparemap; 1840 rxd->rx_size = segs[0].ds_len; 1841 sc->rl_ldata.rl_rx_sparemap = map; 1842 bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap, 1843 BUS_DMASYNC_PREREAD); 1844 1845 desc = &sc->rl_ldata.rl_rx_list[idx]; 1846 desc->rl_vlanctl = 0; 1847 desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[0].ds_addr)); 1848 desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[0].ds_addr)); 1849 cmdstat = segs[0].ds_len; 1850 if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1) 1851 cmdstat |= RL_RDESC_CMD_EOR; 1852 desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN); 1853 1854 return (0); 1855} 1856 1857static int 1858re_jumbo_newbuf(struct rl_softc *sc, int idx) 1859{ 1860 struct mbuf *m; 1861 struct rl_rxdesc *rxd; 1862 bus_dma_segment_t segs[1]; 1863 bus_dmamap_t map; 1864 struct rl_desc *desc; 1865 uint32_t cmdstat; 1866 int error, nsegs; 1867 1868 m = m_getjcl(M_DONTWAIT, MT_DATA, M_PKTHDR, MJUM9BYTES); 1869 if (m == NULL) 1870 return (ENOBUFS); 1871 m->m_len = m->m_pkthdr.len = MJUM9BYTES; 1872#ifdef RE_FIXUP_RX 1873 m_adj(m, RE_ETHER_ALIGN); 1874#endif 1875 error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_jrx_mtag, 1876 sc->rl_ldata.rl_jrx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT); 1877 if (error != 0) { 1878 m_freem(m); 1879 return (ENOBUFS); 1880 } 1881 KASSERT(nsegs == 1, ("%s: %d segment returned!", __func__, nsegs)); 1882 1883 rxd = &sc->rl_ldata.rl_jrx_desc[idx]; 1884 if (rxd->rx_m != NULL) { 1885 bus_dmamap_sync(sc->rl_ldata.rl_jrx_mtag, rxd->rx_dmamap, 1886 BUS_DMASYNC_POSTREAD); 1887 bus_dmamap_unload(sc->rl_ldata.rl_jrx_mtag, rxd->rx_dmamap); 1888 } 1889 1890 rxd->rx_m = m; 1891 map = rxd->rx_dmamap; 1892 rxd->rx_dmamap = sc->rl_ldata.rl_jrx_sparemap; 1893 rxd->rx_size = segs[0].ds_len; 1894 sc->rl_ldata.rl_jrx_sparemap = map; 1895 bus_dmamap_sync(sc->rl_ldata.rl_jrx_mtag, rxd->rx_dmamap, 1896 BUS_DMASYNC_PREREAD); 1897 1898 desc = &sc->rl_ldata.rl_rx_list[idx]; 1899 desc->rl_vlanctl = 0; 1900 desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[0].ds_addr)); 1901 desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[0].ds_addr)); 1902 cmdstat = segs[0].ds_len; 1903 if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1) 1904 cmdstat |= RL_RDESC_CMD_EOR; 1905 desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN); 1906 1907 return (0); 1908} 1909 1910#ifdef RE_FIXUP_RX 1911static __inline void 1912re_fixup_rx(struct mbuf *m) 1913{ 1914 int i; 1915 uint16_t *src, *dst; 1916 1917 src = mtod(m, uint16_t *); 1918 dst = src - (RE_ETHER_ALIGN - ETHER_ALIGN) / sizeof *src; 1919 1920 for (i = 0; i < (m->m_len / sizeof(uint16_t) + 1); i++) 1921 *dst++ = *src++; 1922 1923 m->m_data -= RE_ETHER_ALIGN - ETHER_ALIGN; 1924} 1925#endif 1926 1927static int 1928re_tx_list_init(struct rl_softc *sc) 1929{ 1930 struct rl_desc *desc; 1931 int i; 1932 1933 RL_LOCK_ASSERT(sc); 1934 1935 bzero(sc->rl_ldata.rl_tx_list, 1936 sc->rl_ldata.rl_tx_desc_cnt * sizeof(struct rl_desc)); 1937 for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) 1938 sc->rl_ldata.rl_tx_desc[i].tx_m = NULL; 1939 /* Set EOR. */ 1940 desc = &sc->rl_ldata.rl_tx_list[sc->rl_ldata.rl_tx_desc_cnt - 1]; 1941 desc->rl_cmdstat |= htole32(RL_TDESC_CMD_EOR); 1942 1943 bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag, 1944 sc->rl_ldata.rl_tx_list_map, 1945 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 1946 1947 sc->rl_ldata.rl_tx_prodidx = 0; 1948 sc->rl_ldata.rl_tx_considx = 0; 1949 sc->rl_ldata.rl_tx_free = sc->rl_ldata.rl_tx_desc_cnt; 1950 1951 return (0); 1952} 1953 1954static int 1955re_rx_list_init(struct rl_softc *sc) 1956{ 1957 int error, i; 1958 1959 bzero(sc->rl_ldata.rl_rx_list, 1960 sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc)); 1961 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 1962 sc->rl_ldata.rl_rx_desc[i].rx_m = NULL; 1963 if ((error = re_newbuf(sc, i)) != 0) 1964 return (error); 1965 } 1966 1967 /* Flush the RX descriptors */ 1968 1969 bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag, 1970 sc->rl_ldata.rl_rx_list_map, 1971 BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD); 1972 1973 sc->rl_ldata.rl_rx_prodidx = 0; 1974 sc->rl_head = sc->rl_tail = NULL; 1975 sc->rl_int_rx_act = 0; 1976 1977 return (0); 1978} 1979 1980static int 1981re_jrx_list_init(struct rl_softc *sc) 1982{ 1983 int error, i; 1984 1985 bzero(sc->rl_ldata.rl_rx_list, 1986 sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc)); 1987 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 1988 sc->rl_ldata.rl_jrx_desc[i].rx_m = NULL; 1989 if ((error = re_jumbo_newbuf(sc, i)) != 0) 1990 return (error); 1991 } 1992 1993 bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag, 1994 sc->rl_ldata.rl_rx_list_map, 1995 BUS_DMASYNC_PREWRITE | BUS_DMASYNC_PREREAD); 1996 1997 sc->rl_ldata.rl_rx_prodidx = 0; 1998 sc->rl_head = sc->rl_tail = NULL; 1999 sc->rl_int_rx_act = 0; 2000 2001 return (0); 2002} 2003 2004/* 2005 * RX handler for C+ and 8169. For the gigE chips, we support 2006 * the reception of jumbo frames that have been fragmented 2007 * across multiple 2K mbuf cluster buffers. 2008 */ 2009static int 2010re_rxeof(struct rl_softc *sc, int *rx_npktsp) 2011{ 2012 struct mbuf *m; 2013 struct ifnet *ifp; 2014 int i, rxerr, total_len; 2015 struct rl_desc *cur_rx; 2016 u_int32_t rxstat, rxvlan; 2017 int jumbo, maxpkt = 16, rx_npkts = 0; 2018 2019 RL_LOCK_ASSERT(sc); 2020 2021 ifp = sc->rl_ifp; 2022 if (ifp->if_mtu > RL_MTU && (sc->rl_flags & RL_FLAG_JUMBOV2) != 0) 2023 jumbo = 1; 2024 else 2025 jumbo = 0; 2026 2027 /* Invalidate the descriptor memory */ 2028 2029 bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag, 2030 sc->rl_ldata.rl_rx_list_map, 2031 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 2032 2033 for (i = sc->rl_ldata.rl_rx_prodidx; maxpkt > 0; 2034 i = RL_RX_DESC_NXT(sc, i)) { 2035 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 2036 break; 2037 cur_rx = &sc->rl_ldata.rl_rx_list[i]; 2038 rxstat = le32toh(cur_rx->rl_cmdstat); 2039 if ((rxstat & RL_RDESC_STAT_OWN) != 0) 2040 break; 2041 total_len = rxstat & sc->rl_rxlenmask; 2042 rxvlan = le32toh(cur_rx->rl_vlanctl); 2043 if (jumbo != 0) 2044 m = sc->rl_ldata.rl_jrx_desc[i].rx_m; 2045 else 2046 m = sc->rl_ldata.rl_rx_desc[i].rx_m; 2047 2048 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0 && 2049 (rxstat & (RL_RDESC_STAT_SOF | RL_RDESC_STAT_EOF)) != 2050 (RL_RDESC_STAT_SOF | RL_RDESC_STAT_EOF)) { 2051 /* 2052 * RTL8168C or later controllers do not 2053 * support multi-fragment packet. 2054 */ 2055 re_discard_rxbuf(sc, i); 2056 continue; 2057 } else if ((rxstat & RL_RDESC_STAT_EOF) == 0) { 2058 if (re_newbuf(sc, i) != 0) { 2059 /* 2060 * If this is part of a multi-fragment packet, 2061 * discard all the pieces. 2062 */ 2063 if (sc->rl_head != NULL) { 2064 m_freem(sc->rl_head); 2065 sc->rl_head = sc->rl_tail = NULL; 2066 } 2067 re_discard_rxbuf(sc, i); 2068 continue; 2069 } 2070 m->m_len = RE_RX_DESC_BUFLEN; 2071 if (sc->rl_head == NULL) 2072 sc->rl_head = sc->rl_tail = m; 2073 else { 2074 m->m_flags &= ~M_PKTHDR; 2075 sc->rl_tail->m_next = m; 2076 sc->rl_tail = m; 2077 } 2078 continue; 2079 } 2080 2081 /* 2082 * NOTE: for the 8139C+, the frame length field 2083 * is always 12 bits in size, but for the gigE chips, 2084 * it is 13 bits (since the max RX frame length is 16K). 2085 * Unfortunately, all 32 bits in the status word 2086 * were already used, so to make room for the extra 2087 * length bit, RealTek took out the 'frame alignment 2088 * error' bit and shifted the other status bits 2089 * over one slot. The OWN, EOR, FS and LS bits are 2090 * still in the same places. We have already extracted 2091 * the frame length and checked the OWN bit, so rather 2092 * than using an alternate bit mapping, we shift the 2093 * status bits one space to the right so we can evaluate 2094 * them using the 8169 status as though it was in the 2095 * same format as that of the 8139C+. 2096 */ 2097 if (sc->rl_type == RL_8169) 2098 rxstat >>= 1; 2099 2100 /* 2101 * if total_len > 2^13-1, both _RXERRSUM and _GIANT will be 2102 * set, but if CRC is clear, it will still be a valid frame. 2103 */ 2104 if ((rxstat & RL_RDESC_STAT_RXERRSUM) != 0) { 2105 rxerr = 1; 2106 if ((sc->rl_flags & RL_FLAG_JUMBOV2) == 0 && 2107 total_len > 8191 && 2108 (rxstat & RL_RDESC_STAT_ERRS) == RL_RDESC_STAT_GIANT) 2109 rxerr = 0; 2110 if (rxerr != 0) { 2111 ifp->if_ierrors++; 2112 /* 2113 * If this is part of a multi-fragment packet, 2114 * discard all the pieces. 2115 */ 2116 if (sc->rl_head != NULL) { 2117 m_freem(sc->rl_head); 2118 sc->rl_head = sc->rl_tail = NULL; 2119 } 2120 re_discard_rxbuf(sc, i); 2121 continue; 2122 } 2123 } 2124 2125 /* 2126 * If allocating a replacement mbuf fails, 2127 * reload the current one. 2128 */ 2129 if (jumbo != 0) 2130 rxerr = re_jumbo_newbuf(sc, i); 2131 else 2132 rxerr = re_newbuf(sc, i); 2133 if (rxerr != 0) { 2134 ifp->if_iqdrops++; 2135 if (sc->rl_head != NULL) { 2136 m_freem(sc->rl_head); 2137 sc->rl_head = sc->rl_tail = NULL; 2138 } 2139 re_discard_rxbuf(sc, i); 2140 continue; 2141 } 2142 2143 if (sc->rl_head != NULL) { 2144 if (jumbo != 0) 2145 m->m_len = total_len; 2146 else { 2147 m->m_len = total_len % RE_RX_DESC_BUFLEN; 2148 if (m->m_len == 0) 2149 m->m_len = RE_RX_DESC_BUFLEN; 2150 } 2151 /* 2152 * Special case: if there's 4 bytes or less 2153 * in this buffer, the mbuf can be discarded: 2154 * the last 4 bytes is the CRC, which we don't 2155 * care about anyway. 2156 */ 2157 if (m->m_len <= ETHER_CRC_LEN) { 2158 sc->rl_tail->m_len -= 2159 (ETHER_CRC_LEN - m->m_len); 2160 m_freem(m); 2161 } else { 2162 m->m_len -= ETHER_CRC_LEN; 2163 m->m_flags &= ~M_PKTHDR; 2164 sc->rl_tail->m_next = m; 2165 } 2166 m = sc->rl_head; 2167 sc->rl_head = sc->rl_tail = NULL; 2168 m->m_pkthdr.len = total_len - ETHER_CRC_LEN; 2169 } else 2170 m->m_pkthdr.len = m->m_len = 2171 (total_len - ETHER_CRC_LEN); 2172 2173#ifdef RE_FIXUP_RX 2174 re_fixup_rx(m); 2175#endif 2176 ifp->if_ipackets++; 2177 m->m_pkthdr.rcvif = ifp; 2178 2179 /* Do RX checksumming if enabled */ 2180 2181 if (ifp->if_capenable & IFCAP_RXCSUM) { 2182 if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) { 2183 /* Check IP header checksum */ 2184 if (rxstat & RL_RDESC_STAT_PROTOID) 2185 m->m_pkthdr.csum_flags |= 2186 CSUM_IP_CHECKED; 2187 if (!(rxstat & RL_RDESC_STAT_IPSUMBAD)) 2188 m->m_pkthdr.csum_flags |= 2189 CSUM_IP_VALID; 2190 2191 /* Check TCP/UDP checksum */ 2192 if ((RL_TCPPKT(rxstat) && 2193 !(rxstat & RL_RDESC_STAT_TCPSUMBAD)) || 2194 (RL_UDPPKT(rxstat) && 2195 !(rxstat & RL_RDESC_STAT_UDPSUMBAD))) { 2196 m->m_pkthdr.csum_flags |= 2197 CSUM_DATA_VALID|CSUM_PSEUDO_HDR; 2198 m->m_pkthdr.csum_data = 0xffff; 2199 } 2200 } else { 2201 /* 2202 * RTL8168C/RTL816CP/RTL8111C/RTL8111CP 2203 */ 2204 if ((rxstat & RL_RDESC_STAT_PROTOID) && 2205 (rxvlan & RL_RDESC_IPV4)) 2206 m->m_pkthdr.csum_flags |= 2207 CSUM_IP_CHECKED; 2208 if (!(rxstat & RL_RDESC_STAT_IPSUMBAD) && 2209 (rxvlan & RL_RDESC_IPV4)) 2210 m->m_pkthdr.csum_flags |= 2211 CSUM_IP_VALID; 2212 if (((rxstat & RL_RDESC_STAT_TCP) && 2213 !(rxstat & RL_RDESC_STAT_TCPSUMBAD)) || 2214 ((rxstat & RL_RDESC_STAT_UDP) && 2215 !(rxstat & RL_RDESC_STAT_UDPSUMBAD))) { 2216 m->m_pkthdr.csum_flags |= 2217 CSUM_DATA_VALID|CSUM_PSEUDO_HDR; 2218 m->m_pkthdr.csum_data = 0xffff; 2219 } 2220 } 2221 } 2222 maxpkt--; 2223 if (rxvlan & RL_RDESC_VLANCTL_TAG) { 2224 m->m_pkthdr.ether_vtag = 2225 bswap16((rxvlan & RL_RDESC_VLANCTL_DATA)); 2226 m->m_flags |= M_VLANTAG; 2227 } 2228 RL_UNLOCK(sc); 2229 (*ifp->if_input)(ifp, m); 2230 RL_LOCK(sc); 2231 rx_npkts++; 2232 } 2233 2234 /* Flush the RX DMA ring */ 2235 2236 bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag, 2237 sc->rl_ldata.rl_rx_list_map, 2238 BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD); 2239 2240 sc->rl_ldata.rl_rx_prodidx = i; 2241 2242 if (rx_npktsp != NULL) 2243 *rx_npktsp = rx_npkts; 2244 if (maxpkt) 2245 return (EAGAIN); 2246 2247 return (0); 2248} 2249 2250static void 2251re_txeof(struct rl_softc *sc) 2252{ 2253 struct ifnet *ifp; 2254 struct rl_txdesc *txd; 2255 u_int32_t txstat; 2256 int cons; 2257 2258 cons = sc->rl_ldata.rl_tx_considx; 2259 if (cons == sc->rl_ldata.rl_tx_prodidx) 2260 return; 2261 2262 ifp = sc->rl_ifp; 2263 /* Invalidate the TX descriptor list */ 2264 bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag, 2265 sc->rl_ldata.rl_tx_list_map, 2266 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 2267 2268 for (; cons != sc->rl_ldata.rl_tx_prodidx; 2269 cons = RL_TX_DESC_NXT(sc, cons)) { 2270 txstat = le32toh(sc->rl_ldata.rl_tx_list[cons].rl_cmdstat); 2271 if (txstat & RL_TDESC_STAT_OWN) 2272 break; 2273 /* 2274 * We only stash mbufs in the last descriptor 2275 * in a fragment chain, which also happens to 2276 * be the only place where the TX status bits 2277 * are valid. 2278 */ 2279 if (txstat & RL_TDESC_CMD_EOF) { 2280 txd = &sc->rl_ldata.rl_tx_desc[cons]; 2281 bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag, 2282 txd->tx_dmamap, BUS_DMASYNC_POSTWRITE); 2283 bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag, 2284 txd->tx_dmamap); 2285 KASSERT(txd->tx_m != NULL, 2286 ("%s: freeing NULL mbufs!", __func__)); 2287 m_freem(txd->tx_m); 2288 txd->tx_m = NULL; 2289 if (txstat & (RL_TDESC_STAT_EXCESSCOL| 2290 RL_TDESC_STAT_COLCNT)) 2291 ifp->if_collisions++; 2292 if (txstat & RL_TDESC_STAT_TXERRSUM) 2293 ifp->if_oerrors++; 2294 else 2295 ifp->if_opackets++; 2296 } 2297 sc->rl_ldata.rl_tx_free++; 2298 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 2299 } 2300 sc->rl_ldata.rl_tx_considx = cons; 2301 2302 /* No changes made to the TX ring, so no flush needed */ 2303 2304 if (sc->rl_ldata.rl_tx_free != sc->rl_ldata.rl_tx_desc_cnt) { 2305#ifdef RE_TX_MODERATION 2306 /* 2307 * If not all descriptors have been reaped yet, reload 2308 * the timer so that we will eventually get another 2309 * interrupt that will cause us to re-enter this routine. 2310 * This is done in case the transmitter has gone idle. 2311 */ 2312 CSR_WRITE_4(sc, RL_TIMERCNT, 1); 2313#endif 2314 } else 2315 sc->rl_watchdog_timer = 0; 2316} 2317 2318static void 2319re_tick(void *xsc) 2320{ 2321 struct rl_softc *sc; 2322 struct mii_data *mii; 2323 2324 sc = xsc; 2325 2326 RL_LOCK_ASSERT(sc); 2327 2328 mii = device_get_softc(sc->rl_miibus); 2329 mii_tick(mii); 2330 if ((sc->rl_flags & RL_FLAG_LINK) == 0) 2331 re_miibus_statchg(sc->rl_dev); 2332 /* 2333 * Reclaim transmitted frames here. Technically it is not 2334 * necessary to do here but it ensures periodic reclamation 2335 * regardless of Tx completion interrupt which seems to be 2336 * lost on PCIe based controllers under certain situations. 2337 */ 2338 re_txeof(sc); 2339 re_watchdog(sc); 2340 callout_reset(&sc->rl_stat_callout, hz, re_tick, sc); 2341} 2342 2343#ifdef DEVICE_POLLING 2344static int 2345re_poll(struct ifnet *ifp, enum poll_cmd cmd, int count) 2346{ 2347 struct rl_softc *sc = ifp->if_softc; 2348 int rx_npkts = 0; 2349 2350 RL_LOCK(sc); 2351 if (ifp->if_drv_flags & IFF_DRV_RUNNING) 2352 rx_npkts = re_poll_locked(ifp, cmd, count); 2353 RL_UNLOCK(sc); 2354 return (rx_npkts); 2355} 2356 2357static int 2358re_poll_locked(struct ifnet *ifp, enum poll_cmd cmd, int count) 2359{ 2360 struct rl_softc *sc = ifp->if_softc; 2361 int rx_npkts; 2362 2363 RL_LOCK_ASSERT(sc); 2364 2365 sc->rxcycles = count; 2366 re_rxeof(sc, &rx_npkts); 2367 re_txeof(sc); 2368 2369 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 2370 re_start_locked(ifp); 2371 2372 if (cmd == POLL_AND_CHECK_STATUS) { /* also check status register */ 2373 u_int16_t status; 2374 2375 status = CSR_READ_2(sc, RL_ISR); 2376 if (status == 0xffff) 2377 return (rx_npkts); 2378 if (status) 2379 CSR_WRITE_2(sc, RL_ISR, status); 2380 if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) && 2381 (sc->rl_flags & RL_FLAG_PCIE)) 2382 CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START); 2383 2384 /* 2385 * XXX check behaviour on receiver stalls. 2386 */ 2387 2388 if (status & RL_ISR_SYSTEM_ERR) { 2389 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 2390 re_init_locked(sc); 2391 } 2392 } 2393 return (rx_npkts); 2394} 2395#endif /* DEVICE_POLLING */ 2396 2397static int 2398re_intr(void *arg) 2399{ 2400 struct rl_softc *sc; 2401 uint16_t status; 2402 2403 sc = arg; 2404 2405 status = CSR_READ_2(sc, RL_ISR); 2406 if (status == 0xFFFF || (status & RL_INTRS_CPLUS) == 0) 2407 return (FILTER_STRAY); 2408 CSR_WRITE_2(sc, RL_IMR, 0); 2409 2410 taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_inttask); 2411 2412 return (FILTER_HANDLED); 2413} 2414 2415static void 2416re_int_task(void *arg, int npending) 2417{ 2418 struct rl_softc *sc; 2419 struct ifnet *ifp; 2420 u_int16_t status; 2421 int rval = 0; 2422 2423 sc = arg; 2424 ifp = sc->rl_ifp; 2425 2426 RL_LOCK(sc); 2427 2428 status = CSR_READ_2(sc, RL_ISR); 2429 CSR_WRITE_2(sc, RL_ISR, status); 2430 2431 if (sc->suspended || 2432 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 2433 RL_UNLOCK(sc); 2434 return; 2435 } 2436 2437#ifdef DEVICE_POLLING 2438 if (ifp->if_capenable & IFCAP_POLLING) { 2439 RL_UNLOCK(sc); 2440 return; 2441 } 2442#endif 2443 2444 if (status & (RL_ISR_RX_OK|RL_ISR_RX_ERR|RL_ISR_FIFO_OFLOW)) 2445 rval = re_rxeof(sc, NULL); 2446 2447 /* 2448 * Some chips will ignore a second TX request issued 2449 * while an existing transmission is in progress. If 2450 * the transmitter goes idle but there are still 2451 * packets waiting to be sent, we need to restart the 2452 * channel here to flush them out. This only seems to 2453 * be required with the PCIe devices. 2454 */ 2455 if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) && 2456 (sc->rl_flags & RL_FLAG_PCIE)) 2457 CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START); 2458 if (status & ( 2459#ifdef RE_TX_MODERATION 2460 RL_ISR_TIMEOUT_EXPIRED| 2461#else 2462 RL_ISR_TX_OK| 2463#endif 2464 RL_ISR_TX_ERR|RL_ISR_TX_DESC_UNAVAIL)) 2465 re_txeof(sc); 2466 2467 if (status & RL_ISR_SYSTEM_ERR) { 2468 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 2469 re_init_locked(sc); 2470 } 2471 2472 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 2473 re_start_locked(ifp); 2474 2475 RL_UNLOCK(sc); 2476 2477 if ((CSR_READ_2(sc, RL_ISR) & RL_INTRS_CPLUS) || rval) { 2478 taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_inttask); 2479 return; 2480 } 2481 2482 CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS); 2483} 2484 2485static void 2486re_intr_msi(void *xsc) 2487{ 2488 struct rl_softc *sc; 2489 struct ifnet *ifp; 2490 uint16_t intrs, status; 2491 2492 sc = xsc; 2493 RL_LOCK(sc); 2494 2495 ifp = sc->rl_ifp; 2496#ifdef DEVICE_POLLING 2497 if (ifp->if_capenable & IFCAP_POLLING) { 2498 RL_UNLOCK(sc); 2499 return; 2500 } 2501#endif 2502 /* Disable interrupts. */ 2503 CSR_WRITE_2(sc, RL_IMR, 0); 2504 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 2505 RL_UNLOCK(sc); 2506 return; 2507 } 2508 2509 intrs = RL_INTRS_CPLUS; 2510 status = CSR_READ_2(sc, RL_ISR); 2511 CSR_WRITE_2(sc, RL_ISR, status); 2512 if (sc->rl_int_rx_act > 0) { 2513 intrs &= ~(RL_ISR_RX_OK | RL_ISR_RX_ERR | RL_ISR_FIFO_OFLOW | 2514 RL_ISR_RX_OVERRUN); 2515 status &= ~(RL_ISR_RX_OK | RL_ISR_RX_ERR | RL_ISR_FIFO_OFLOW | 2516 RL_ISR_RX_OVERRUN); 2517 } 2518 2519 if (status & (RL_ISR_TIMEOUT_EXPIRED | RL_ISR_RX_OK | RL_ISR_RX_ERR | 2520 RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN)) { 2521 re_rxeof(sc, NULL); 2522 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) { 2523 if (sc->rl_int_rx_mod != 0 && 2524 (status & (RL_ISR_RX_OK | RL_ISR_RX_ERR | 2525 RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN)) != 0) { 2526 /* Rearm one-shot timer. */ 2527 CSR_WRITE_4(sc, RL_TIMERCNT, 1); 2528 intrs &= ~(RL_ISR_RX_OK | RL_ISR_RX_ERR | 2529 RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN); 2530 sc->rl_int_rx_act = 1; 2531 } else { 2532 intrs |= RL_ISR_RX_OK | RL_ISR_RX_ERR | 2533 RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN; 2534 sc->rl_int_rx_act = 0; 2535 } 2536 } 2537 } 2538 2539 /* 2540 * Some chips will ignore a second TX request issued 2541 * while an existing transmission is in progress. If 2542 * the transmitter goes idle but there are still 2543 * packets waiting to be sent, we need to restart the 2544 * channel here to flush them out. This only seems to 2545 * be required with the PCIe devices. 2546 */ 2547 if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) && 2548 (sc->rl_flags & RL_FLAG_PCIE)) 2549 CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START); 2550 if (status & (RL_ISR_TX_OK | RL_ISR_TX_ERR | RL_ISR_TX_DESC_UNAVAIL)) 2551 re_txeof(sc); 2552 2553 if (status & RL_ISR_SYSTEM_ERR) { 2554 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 2555 re_init_locked(sc); 2556 } 2557 2558 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) { 2559 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 2560 re_start_locked(ifp); 2561 CSR_WRITE_2(sc, RL_IMR, intrs); 2562 } 2563 RL_UNLOCK(sc); 2564} 2565 2566static int 2567re_encap(struct rl_softc *sc, struct mbuf **m_head) 2568{ 2569 struct rl_txdesc *txd, *txd_last; 2570 bus_dma_segment_t segs[RL_NTXSEGS]; 2571 bus_dmamap_t map; 2572 struct mbuf *m_new; 2573 struct rl_desc *desc; 2574 int nsegs, prod; 2575 int i, error, ei, si; 2576 int padlen; 2577 uint32_t cmdstat, csum_flags, vlanctl; 2578 2579 RL_LOCK_ASSERT(sc); 2580 M_ASSERTPKTHDR((*m_head)); 2581 2582 /* 2583 * With some of the RealTek chips, using the checksum offload 2584 * support in conjunction with the autopadding feature results 2585 * in the transmission of corrupt frames. For example, if we 2586 * need to send a really small IP fragment that's less than 60 2587 * bytes in size, and IP header checksumming is enabled, the 2588 * resulting ethernet frame that appears on the wire will 2589 * have garbled payload. To work around this, if TX IP checksum 2590 * offload is enabled, we always manually pad short frames out 2591 * to the minimum ethernet frame size. 2592 */ 2593 if ((sc->rl_flags & RL_FLAG_AUTOPAD) == 0 && 2594 (*m_head)->m_pkthdr.len < RL_IP4CSUMTX_PADLEN && 2595 ((*m_head)->m_pkthdr.csum_flags & CSUM_IP) != 0) { 2596 padlen = RL_MIN_FRAMELEN - (*m_head)->m_pkthdr.len; 2597 if (M_WRITABLE(*m_head) == 0) { 2598 /* Get a writable copy. */ 2599 m_new = m_dup(*m_head, M_DONTWAIT); 2600 m_freem(*m_head); 2601 if (m_new == NULL) { 2602 *m_head = NULL; 2603 return (ENOBUFS); 2604 } 2605 *m_head = m_new; 2606 } 2607 if ((*m_head)->m_next != NULL || 2608 M_TRAILINGSPACE(*m_head) < padlen) { 2609 m_new = m_defrag(*m_head, M_DONTWAIT); 2610 if (m_new == NULL) { 2611 m_freem(*m_head); 2612 *m_head = NULL; 2613 return (ENOBUFS); 2614 } 2615 } else 2616 m_new = *m_head; 2617 2618 /* 2619 * Manually pad short frames, and zero the pad space 2620 * to avoid leaking data. 2621 */ 2622 bzero(mtod(m_new, char *) + m_new->m_pkthdr.len, padlen); 2623 m_new->m_pkthdr.len += padlen; 2624 m_new->m_len = m_new->m_pkthdr.len; 2625 *m_head = m_new; 2626 } 2627 2628 prod = sc->rl_ldata.rl_tx_prodidx; 2629 txd = &sc->rl_ldata.rl_tx_desc[prod]; 2630 error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap, 2631 *m_head, segs, &nsegs, BUS_DMA_NOWAIT); 2632 if (error == EFBIG) { 2633 m_new = m_collapse(*m_head, M_DONTWAIT, RL_NTXSEGS); 2634 if (m_new == NULL) { 2635 m_freem(*m_head); 2636 *m_head = NULL; 2637 return (ENOBUFS); 2638 } 2639 *m_head = m_new; 2640 error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_tx_mtag, 2641 txd->tx_dmamap, *m_head, segs, &nsegs, BUS_DMA_NOWAIT); 2642 if (error != 0) { 2643 m_freem(*m_head); 2644 *m_head = NULL; 2645 return (error); 2646 } 2647 } else if (error != 0) 2648 return (error); 2649 if (nsegs == 0) { 2650 m_freem(*m_head); 2651 *m_head = NULL; 2652 return (EIO); 2653 } 2654 2655 /* Check for number of available descriptors. */ 2656 if (sc->rl_ldata.rl_tx_free - nsegs <= 1) { 2657 bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap); 2658 return (ENOBUFS); 2659 } 2660 2661 bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap, 2662 BUS_DMASYNC_PREWRITE); 2663 2664 /* 2665 * Set up checksum offload. Note: checksum offload bits must 2666 * appear in all descriptors of a multi-descriptor transmit 2667 * attempt. This is according to testing done with an 8169 2668 * chip. This is a requirement. 2669 */ 2670 vlanctl = 0; 2671 csum_flags = 0; 2672 if (((*m_head)->m_pkthdr.csum_flags & CSUM_TSO) != 0) { 2673 if ((sc->rl_flags & RL_FLAG_DESCV2) != 0) { 2674 csum_flags |= RL_TDESC_CMD_LGSEND; 2675 vlanctl |= ((uint32_t)(*m_head)->m_pkthdr.tso_segsz << 2676 RL_TDESC_CMD_MSSVALV2_SHIFT); 2677 } else { 2678 csum_flags |= RL_TDESC_CMD_LGSEND | 2679 ((uint32_t)(*m_head)->m_pkthdr.tso_segsz << 2680 RL_TDESC_CMD_MSSVAL_SHIFT); 2681 } 2682 } else { 2683 /* 2684 * Unconditionally enable IP checksum if TCP or UDP 2685 * checksum is required. Otherwise, TCP/UDP checksum 2686 * does't make effects. 2687 */ 2688 if (((*m_head)->m_pkthdr.csum_flags & RE_CSUM_FEATURES) != 0) { 2689 if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) { 2690 csum_flags |= RL_TDESC_CMD_IPCSUM; 2691 if (((*m_head)->m_pkthdr.csum_flags & 2692 CSUM_TCP) != 0) 2693 csum_flags |= RL_TDESC_CMD_TCPCSUM; 2694 if (((*m_head)->m_pkthdr.csum_flags & 2695 CSUM_UDP) != 0) 2696 csum_flags |= RL_TDESC_CMD_UDPCSUM; 2697 } else { 2698 vlanctl |= RL_TDESC_CMD_IPCSUMV2; 2699 if (((*m_head)->m_pkthdr.csum_flags & 2700 CSUM_TCP) != 0) 2701 vlanctl |= RL_TDESC_CMD_TCPCSUMV2; 2702 if (((*m_head)->m_pkthdr.csum_flags & 2703 CSUM_UDP) != 0) 2704 vlanctl |= RL_TDESC_CMD_UDPCSUMV2; 2705 } 2706 } 2707 } 2708 2709 /* 2710 * Set up hardware VLAN tagging. Note: vlan tag info must 2711 * appear in all descriptors of a multi-descriptor 2712 * transmission attempt. 2713 */ 2714 if ((*m_head)->m_flags & M_VLANTAG) 2715 vlanctl |= bswap16((*m_head)->m_pkthdr.ether_vtag) | 2716 RL_TDESC_VLANCTL_TAG; 2717 2718 si = prod; 2719 for (i = 0; i < nsegs; i++, prod = RL_TX_DESC_NXT(sc, prod)) { 2720 desc = &sc->rl_ldata.rl_tx_list[prod]; 2721 desc->rl_vlanctl = htole32(vlanctl); 2722 desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[i].ds_addr)); 2723 desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[i].ds_addr)); 2724 cmdstat = segs[i].ds_len; 2725 if (i != 0) 2726 cmdstat |= RL_TDESC_CMD_OWN; 2727 if (prod == sc->rl_ldata.rl_tx_desc_cnt - 1) 2728 cmdstat |= RL_TDESC_CMD_EOR; 2729 desc->rl_cmdstat = htole32(cmdstat | csum_flags); 2730 sc->rl_ldata.rl_tx_free--; 2731 } 2732 /* Update producer index. */ 2733 sc->rl_ldata.rl_tx_prodidx = prod; 2734 2735 /* Set EOF on the last descriptor. */ 2736 ei = RL_TX_DESC_PRV(sc, prod); 2737 desc = &sc->rl_ldata.rl_tx_list[ei]; 2738 desc->rl_cmdstat |= htole32(RL_TDESC_CMD_EOF); 2739 2740 desc = &sc->rl_ldata.rl_tx_list[si]; 2741 /* Set SOF and transfer ownership of packet to the chip. */ 2742 desc->rl_cmdstat |= htole32(RL_TDESC_CMD_OWN | RL_TDESC_CMD_SOF); 2743 2744 /* 2745 * Insure that the map for this transmission 2746 * is placed at the array index of the last descriptor 2747 * in this chain. (Swap last and first dmamaps.) 2748 */ 2749 txd_last = &sc->rl_ldata.rl_tx_desc[ei]; 2750 map = txd->tx_dmamap; 2751 txd->tx_dmamap = txd_last->tx_dmamap; 2752 txd_last->tx_dmamap = map; 2753 txd_last->tx_m = *m_head; 2754 2755 return (0); 2756} 2757 2758static void 2759re_start(struct ifnet *ifp) 2760{ 2761 struct rl_softc *sc; 2762 2763 sc = ifp->if_softc; 2764 RL_LOCK(sc); 2765 re_start_locked(ifp); 2766 RL_UNLOCK(sc); 2767} 2768 2769/* 2770 * Main transmit routine for C+ and gigE NICs. 2771 */ 2772static void 2773re_start_locked(struct ifnet *ifp) 2774{ 2775 struct rl_softc *sc; 2776 struct mbuf *m_head; 2777 int queued; 2778 2779 sc = ifp->if_softc; 2780 2781 if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) != 2782 IFF_DRV_RUNNING || (sc->rl_flags & RL_FLAG_LINK) == 0) 2783 return; 2784 2785 for (queued = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd) && 2786 sc->rl_ldata.rl_tx_free > 1;) { 2787 IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head); 2788 if (m_head == NULL) 2789 break; 2790 2791 if (re_encap(sc, &m_head) != 0) { 2792 if (m_head == NULL) 2793 break; 2794 IFQ_DRV_PREPEND(&ifp->if_snd, m_head); 2795 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 2796 break; 2797 } 2798 2799 /* 2800 * If there's a BPF listener, bounce a copy of this frame 2801 * to him. 2802 */ 2803 ETHER_BPF_MTAP(ifp, m_head); 2804 2805 queued++; 2806 } 2807 2808 if (queued == 0) { 2809#ifdef RE_TX_MODERATION 2810 if (sc->rl_ldata.rl_tx_free != sc->rl_ldata.rl_tx_desc_cnt) 2811 CSR_WRITE_4(sc, RL_TIMERCNT, 1); 2812#endif 2813 return; 2814 } 2815 2816 /* Flush the TX descriptors */ 2817 2818 bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag, 2819 sc->rl_ldata.rl_tx_list_map, 2820 BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD); 2821 2822 CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START); 2823 2824#ifdef RE_TX_MODERATION 2825 /* 2826 * Use the countdown timer for interrupt moderation. 2827 * 'TX done' interrupts are disabled. Instead, we reset the 2828 * countdown timer, which will begin counting until it hits 2829 * the value in the TIMERINT register, and then trigger an 2830 * interrupt. Each time we write to the TIMERCNT register, 2831 * the timer count is reset to 0. 2832 */ 2833 CSR_WRITE_4(sc, RL_TIMERCNT, 1); 2834#endif 2835 2836 /* 2837 * Set a timeout in case the chip goes out to lunch. 2838 */ 2839 sc->rl_watchdog_timer = 5; 2840} 2841 2842static void 2843re_set_jumbo(struct rl_softc *sc, int jumbo) 2844{ 2845 2846 if (sc->rl_hwrev->rl_rev == RL_HWREV_8168E_VL) { 2847 pci_set_max_read_req(sc->rl_dev, 4096); 2848 return; 2849 } 2850 2851 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_WRITECFG); 2852 if (jumbo != 0) { 2853 CSR_WRITE_1(sc, RL_CFG3, CSR_READ_1(sc, RL_CFG3) | 2854 RL_CFG3_JUMBO_EN0); 2855 switch (sc->rl_hwrev->rl_rev) { 2856 case RL_HWREV_8168DP: 2857 break; 2858 case RL_HWREV_8168E: 2859 CSR_WRITE_1(sc, RL_CFG4, CSR_READ_1(sc, RL_CFG4) | 2860 0x01); 2861 break; 2862 default: 2863 CSR_WRITE_1(sc, RL_CFG4, CSR_READ_1(sc, RL_CFG4) | 2864 RL_CFG4_JUMBO_EN1); 2865 } 2866 } else { 2867 CSR_WRITE_1(sc, RL_CFG3, CSR_READ_1(sc, RL_CFG3) & 2868 ~RL_CFG3_JUMBO_EN0); 2869 switch (sc->rl_hwrev->rl_rev) { 2870 case RL_HWREV_8168DP: 2871 break; 2872 case RL_HWREV_8168E: 2873 CSR_WRITE_1(sc, RL_CFG4, CSR_READ_1(sc, RL_CFG4) & 2874 ~0x01); 2875 break; 2876 default: 2877 CSR_WRITE_1(sc, RL_CFG4, CSR_READ_1(sc, RL_CFG4) & 2878 ~RL_CFG4_JUMBO_EN1); 2879 } 2880 } 2881 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 2882 2883 switch (sc->rl_hwrev->rl_rev) { 2884 case RL_HWREV_8168DP: 2885 pci_set_max_read_req(sc->rl_dev, 4096); 2886 break; 2887 default: 2888 if (jumbo != 0) 2889 pci_set_max_read_req(sc->rl_dev, 512); 2890 else 2891 pci_set_max_read_req(sc->rl_dev, 4096); 2892 } 2893} 2894 2895static void 2896re_init(void *xsc) 2897{ 2898 struct rl_softc *sc = xsc; 2899 2900 RL_LOCK(sc); 2901 re_init_locked(sc); 2902 RL_UNLOCK(sc); 2903} 2904 2905static void 2906re_init_locked(struct rl_softc *sc) 2907{ 2908 struct ifnet *ifp = sc->rl_ifp; 2909 struct mii_data *mii; 2910 uint32_t reg; 2911 uint16_t cfg; 2912 union { 2913 uint32_t align_dummy; 2914 u_char eaddr[ETHER_ADDR_LEN]; 2915 } eaddr; 2916 2917 RL_LOCK_ASSERT(sc); 2918 2919 mii = device_get_softc(sc->rl_miibus); 2920 2921 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) 2922 return; 2923 2924 /* 2925 * Cancel pending I/O and free all RX/TX buffers. 2926 */ 2927 re_stop(sc); 2928 2929 /* Put controller into known state. */ 2930 re_reset(sc); 2931 2932 /* 2933 * For C+ mode, initialize the RX descriptors and mbufs. 2934 */ 2935 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) { 2936 if (ifp->if_mtu > RL_MTU) { 2937 if (re_jrx_list_init(sc) != 0) { 2938 device_printf(sc->rl_dev, 2939 "no memory for jumbo RX buffers\n"); 2940 re_stop(sc); 2941 return; 2942 } 2943 /* Disable checksum offloading for jumbo frames. */ 2944 ifp->if_capenable &= ~(IFCAP_HWCSUM | IFCAP_TSO4); 2945 ifp->if_hwassist &= ~(RE_CSUM_FEATURES | CSUM_TSO); 2946 } else { 2947 if (re_rx_list_init(sc) != 0) { 2948 device_printf(sc->rl_dev, 2949 "no memory for RX buffers\n"); 2950 re_stop(sc); 2951 return; 2952 } 2953 } 2954 re_set_jumbo(sc, ifp->if_mtu > RL_MTU); 2955 } else { 2956 if (re_rx_list_init(sc) != 0) { 2957 device_printf(sc->rl_dev, "no memory for RX buffers\n"); 2958 re_stop(sc); 2959 return; 2960 } 2961 if ((sc->rl_flags & RL_FLAG_PCIE) != 0 && 2962 pci_get_device(sc->rl_dev) != RT_DEVICEID_8101E) { 2963 if (ifp->if_mtu > RL_MTU) 2964 pci_set_max_read_req(sc->rl_dev, 512); 2965 else 2966 pci_set_max_read_req(sc->rl_dev, 4096); 2967 } 2968 } 2969 re_tx_list_init(sc); 2970 2971 /* 2972 * Enable C+ RX and TX mode, as well as VLAN stripping and 2973 * RX checksum offload. We must configure the C+ register 2974 * before all others. 2975 */ 2976 cfg = RL_CPLUSCMD_PCI_MRW; 2977 if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) 2978 cfg |= RL_CPLUSCMD_RXCSUM_ENB; 2979 if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) 2980 cfg |= RL_CPLUSCMD_VLANSTRIP; 2981 if ((sc->rl_flags & RL_FLAG_MACSTAT) != 0) { 2982 cfg |= RL_CPLUSCMD_MACSTAT_DIS; 2983 /* XXX magic. */ 2984 cfg |= 0x0001; 2985 } else 2986 cfg |= RL_CPLUSCMD_RXENB | RL_CPLUSCMD_TXENB; 2987 CSR_WRITE_2(sc, RL_CPLUS_CMD, cfg); 2988 if (sc->rl_hwrev->rl_rev == RL_HWREV_8169_8110SC || 2989 sc->rl_hwrev->rl_rev == RL_HWREV_8169_8110SCE) { 2990 reg = 0x000fff00; 2991 if ((CSR_READ_1(sc, RL_CFG2) & RL_CFG2_PCI66MHZ) != 0) 2992 reg |= 0x000000ff; 2993 if (sc->rl_hwrev->rl_rev == RL_HWREV_8169_8110SCE) 2994 reg |= 0x00f00000; 2995 CSR_WRITE_4(sc, 0x7c, reg); 2996 /* Disable interrupt mitigation. */ 2997 CSR_WRITE_2(sc, 0xe2, 0); 2998 } 2999 /* 3000 * Disable TSO if interface MTU size is greater than MSS 3001 * allowed in controller. 3002 */ 3003 if (ifp->if_mtu > RL_TSO_MTU && (ifp->if_capenable & IFCAP_TSO4) != 0) { 3004 ifp->if_capenable &= ~IFCAP_TSO4; 3005 ifp->if_hwassist &= ~CSUM_TSO; 3006 } 3007 3008 /* 3009 * Init our MAC address. Even though the chipset 3010 * documentation doesn't mention it, we need to enter "Config 3011 * register write enable" mode to modify the ID registers. 3012 */ 3013 /* Copy MAC address on stack to align. */ 3014 bcopy(IF_LLADDR(ifp), eaddr.eaddr, ETHER_ADDR_LEN); 3015 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_WRITECFG); 3016 CSR_WRITE_4(sc, RL_IDR0, 3017 htole32(*(u_int32_t *)(&eaddr.eaddr[0]))); 3018 CSR_WRITE_4(sc, RL_IDR4, 3019 htole32(*(u_int32_t *)(&eaddr.eaddr[4]))); 3020 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 3021 3022 /* 3023 * Load the addresses of the RX and TX lists into the chip. 3024 */ 3025 3026 CSR_WRITE_4(sc, RL_RXLIST_ADDR_HI, 3027 RL_ADDR_HI(sc->rl_ldata.rl_rx_list_addr)); 3028 CSR_WRITE_4(sc, RL_RXLIST_ADDR_LO, 3029 RL_ADDR_LO(sc->rl_ldata.rl_rx_list_addr)); 3030 3031 CSR_WRITE_4(sc, RL_TXLIST_ADDR_HI, 3032 RL_ADDR_HI(sc->rl_ldata.rl_tx_list_addr)); 3033 CSR_WRITE_4(sc, RL_TXLIST_ADDR_LO, 3034 RL_ADDR_LO(sc->rl_ldata.rl_tx_list_addr)); 3035 3036 /* 3037 * Enable transmit and receive. 3038 */ 3039 CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_TX_ENB|RL_CMD_RX_ENB); 3040 3041 /* 3042 * Set the initial TX configuration. 3043 */ 3044 if (sc->rl_testmode) { 3045 if (sc->rl_type == RL_8169) 3046 CSR_WRITE_4(sc, RL_TXCFG, 3047 RL_TXCFG_CONFIG|RL_LOOPTEST_ON); 3048 else 3049 CSR_WRITE_4(sc, RL_TXCFG, 3050 RL_TXCFG_CONFIG|RL_LOOPTEST_ON_CPLUS); 3051 } else 3052 CSR_WRITE_4(sc, RL_TXCFG, RL_TXCFG_CONFIG); 3053 3054 CSR_WRITE_1(sc, RL_EARLY_TX_THRESH, 16); 3055 3056 /* 3057 * Set the initial RX configuration. 3058 */ 3059 re_set_rxmode(sc); 3060 3061 /* Configure interrupt moderation. */ 3062 if (sc->rl_type == RL_8169) { 3063 /* Magic from vendor. */ 3064 CSR_WRITE_2(sc, RL_INTRMOD, 0x5100); 3065 } 3066 3067#ifdef DEVICE_POLLING 3068 /* 3069 * Disable interrupts if we are polling. 3070 */ 3071 if (ifp->if_capenable & IFCAP_POLLING) 3072 CSR_WRITE_2(sc, RL_IMR, 0); 3073 else /* otherwise ... */ 3074#endif 3075 3076 /* 3077 * Enable interrupts. 3078 */ 3079 if (sc->rl_testmode) 3080 CSR_WRITE_2(sc, RL_IMR, 0); 3081 else 3082 CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS); 3083 CSR_WRITE_2(sc, RL_ISR, RL_INTRS_CPLUS); 3084 3085 /* Set initial TX threshold */ 3086 sc->rl_txthresh = RL_TX_THRESH_INIT; 3087 3088 /* Start RX/TX process. */ 3089 CSR_WRITE_4(sc, RL_MISSEDPKT, 0); 3090#ifdef notdef 3091 /* Enable receiver and transmitter. */ 3092 CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_TX_ENB|RL_CMD_RX_ENB); 3093#endif 3094 3095 /* 3096 * Initialize the timer interrupt register so that 3097 * a timer interrupt will be generated once the timer 3098 * reaches a certain number of ticks. The timer is 3099 * reloaded on each transmit. 3100 */ 3101#ifdef RE_TX_MODERATION 3102 /* 3103 * Use timer interrupt register to moderate TX interrupt 3104 * moderation, which dramatically improves TX frame rate. 3105 */ 3106 if (sc->rl_type == RL_8169) 3107 CSR_WRITE_4(sc, RL_TIMERINT_8169, 0x800); 3108 else 3109 CSR_WRITE_4(sc, RL_TIMERINT, 0x400); 3110#else 3111 /* 3112 * Use timer interrupt register to moderate RX interrupt 3113 * moderation. 3114 */ 3115 if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) != 0 && 3116 intr_filter == 0) { 3117 if (sc->rl_type == RL_8169) 3118 CSR_WRITE_4(sc, RL_TIMERINT_8169, 3119 RL_USECS(sc->rl_int_rx_mod)); 3120 } else { 3121 if (sc->rl_type == RL_8169) 3122 CSR_WRITE_4(sc, RL_TIMERINT_8169, RL_USECS(0)); 3123 } 3124#endif 3125 3126 /* 3127 * For 8169 gigE NICs, set the max allowed RX packet 3128 * size so we can receive jumbo frames. 3129 */ 3130 if (sc->rl_type == RL_8169) { 3131 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) { 3132 /* 3133 * For controllers that use new jumbo frame scheme, 3134 * set maximum size of jumbo frame depedning on 3135 * controller revisions. 3136 */ 3137 if (ifp->if_mtu > RL_MTU) 3138 CSR_WRITE_2(sc, RL_MAXRXPKTLEN, 3139 sc->rl_hwrev->rl_max_mtu + 3140 ETHER_VLAN_ENCAP_LEN + ETHER_HDR_LEN + 3141 ETHER_CRC_LEN); 3142 else 3143 CSR_WRITE_2(sc, RL_MAXRXPKTLEN, 3144 RE_RX_DESC_BUFLEN); 3145 } else if ((sc->rl_flags & RL_FLAG_PCIE) != 0 && 3146 sc->rl_hwrev->rl_max_mtu == RL_MTU) { 3147 /* RTL810x has no jumbo frame support. */ 3148 CSR_WRITE_2(sc, RL_MAXRXPKTLEN, RE_RX_DESC_BUFLEN); 3149 } else 3150 CSR_WRITE_2(sc, RL_MAXRXPKTLEN, 16383); 3151 } 3152 3153 if (sc->rl_testmode) 3154 return; 3155 3156 mii_mediachg(mii); 3157 3158 CSR_WRITE_1(sc, RL_CFG1, CSR_READ_1(sc, RL_CFG1) | RL_CFG1_DRVLOAD); 3159 3160 ifp->if_drv_flags |= IFF_DRV_RUNNING; 3161 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 3162 3163 sc->rl_flags &= ~RL_FLAG_LINK; 3164 sc->rl_watchdog_timer = 0; 3165 callout_reset(&sc->rl_stat_callout, hz, re_tick, sc); 3166} 3167 3168/* 3169 * Set media options. 3170 */ 3171static int 3172re_ifmedia_upd(struct ifnet *ifp) 3173{ 3174 struct rl_softc *sc; 3175 struct mii_data *mii; 3176 int error; 3177 3178 sc = ifp->if_softc; 3179 mii = device_get_softc(sc->rl_miibus); 3180 RL_LOCK(sc); 3181 error = mii_mediachg(mii); 3182 RL_UNLOCK(sc); 3183 3184 return (error); 3185} 3186 3187/* 3188 * Report current media status. 3189 */ 3190static void 3191re_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr) 3192{ 3193 struct rl_softc *sc; 3194 struct mii_data *mii; 3195 3196 sc = ifp->if_softc; 3197 mii = device_get_softc(sc->rl_miibus); 3198 3199 RL_LOCK(sc); 3200 mii_pollstat(mii); 3201 RL_UNLOCK(sc); 3202 ifmr->ifm_active = mii->mii_media_active; 3203 ifmr->ifm_status = mii->mii_media_status; 3204} 3205 3206static int 3207re_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 3208{ 3209 struct rl_softc *sc = ifp->if_softc; 3210 struct ifreq *ifr = (struct ifreq *) data; 3211 struct mii_data *mii; 3212 int error = 0; 3213 3214 switch (command) { 3215 case SIOCSIFMTU: 3216 if (ifr->ifr_mtu < ETHERMIN || 3217 ifr->ifr_mtu > sc->rl_hwrev->rl_max_mtu) { 3218 error = EINVAL; 3219 break; 3220 } 3221 RL_LOCK(sc); 3222 if (ifp->if_mtu != ifr->ifr_mtu) { 3223 ifp->if_mtu = ifr->ifr_mtu; 3224 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0 && 3225 (ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) { 3226 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 3227 re_init_locked(sc); 3228 } 3229 if (ifp->if_mtu > RL_TSO_MTU && 3230 (ifp->if_capenable & IFCAP_TSO4) != 0) { 3231 ifp->if_capenable &= ~(IFCAP_TSO4 | 3232 IFCAP_VLAN_HWTSO); 3233 ifp->if_hwassist &= ~CSUM_TSO; 3234 } 3235 VLAN_CAPABILITIES(ifp); 3236 } 3237 RL_UNLOCK(sc); 3238 break; 3239 case SIOCSIFFLAGS: 3240 RL_LOCK(sc); 3241 if ((ifp->if_flags & IFF_UP) != 0) { 3242 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) { 3243 if (((ifp->if_flags ^ sc->rl_if_flags) 3244 & (IFF_PROMISC | IFF_ALLMULTI)) != 0) 3245 re_set_rxmode(sc); 3246 } else 3247 re_init_locked(sc); 3248 } else { 3249 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) 3250 re_stop(sc); 3251 } 3252 sc->rl_if_flags = ifp->if_flags; 3253 RL_UNLOCK(sc); 3254 break; 3255 case SIOCADDMULTI: 3256 case SIOCDELMULTI: 3257 RL_LOCK(sc); 3258 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) 3259 re_set_rxmode(sc); 3260 RL_UNLOCK(sc); 3261 break; 3262 case SIOCGIFMEDIA: 3263 case SIOCSIFMEDIA: 3264 mii = device_get_softc(sc->rl_miibus); 3265 error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command); 3266 break; 3267 case SIOCSIFCAP: 3268 { 3269 int mask, reinit; 3270 3271 mask = ifr->ifr_reqcap ^ ifp->if_capenable; 3272 reinit = 0; 3273#ifdef DEVICE_POLLING 3274 if (mask & IFCAP_POLLING) { 3275 if (ifr->ifr_reqcap & IFCAP_POLLING) { 3276 error = ether_poll_register(re_poll, ifp); 3277 if (error) 3278 return (error); 3279 RL_LOCK(sc); 3280 /* Disable interrupts */ 3281 CSR_WRITE_2(sc, RL_IMR, 0x0000); 3282 ifp->if_capenable |= IFCAP_POLLING; 3283 RL_UNLOCK(sc); 3284 } else { 3285 error = ether_poll_deregister(ifp); 3286 /* Enable interrupts. */ 3287 RL_LOCK(sc); 3288 CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS); 3289 ifp->if_capenable &= ~IFCAP_POLLING; 3290 RL_UNLOCK(sc); 3291 } 3292 } 3293#endif /* DEVICE_POLLING */ 3294 if ((mask & IFCAP_TXCSUM) != 0 && 3295 (ifp->if_capabilities & IFCAP_TXCSUM) != 0) { 3296 ifp->if_capenable ^= IFCAP_TXCSUM; 3297 if ((ifp->if_capenable & IFCAP_TXCSUM) != 0) 3298 ifp->if_hwassist |= RE_CSUM_FEATURES; 3299 else 3300 ifp->if_hwassist &= ~RE_CSUM_FEATURES; 3301 reinit = 1; 3302 } 3303 if ((mask & IFCAP_RXCSUM) != 0 && 3304 (ifp->if_capabilities & IFCAP_RXCSUM) != 0) { 3305 ifp->if_capenable ^= IFCAP_RXCSUM; 3306 reinit = 1; 3307 } 3308 if ((mask & IFCAP_TSO4) != 0 && 3309 (ifp->if_capabilities & IFCAP_TSO) != 0) { 3310 ifp->if_capenable ^= IFCAP_TSO4; 3311 if ((IFCAP_TSO4 & ifp->if_capenable) != 0) 3312 ifp->if_hwassist |= CSUM_TSO; 3313 else 3314 ifp->if_hwassist &= ~CSUM_TSO; 3315 if (ifp->if_mtu > RL_TSO_MTU && 3316 (ifp->if_capenable & IFCAP_TSO4) != 0) { 3317 ifp->if_capenable &= ~IFCAP_TSO4; 3318 ifp->if_hwassist &= ~CSUM_TSO; 3319 } 3320 } 3321 if ((mask & IFCAP_VLAN_HWTSO) != 0 && 3322 (ifp->if_capabilities & IFCAP_VLAN_HWTSO) != 0) 3323 ifp->if_capenable ^= IFCAP_VLAN_HWTSO; 3324 if ((mask & IFCAP_VLAN_HWTAGGING) != 0 && 3325 (ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0) { 3326 ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; 3327 /* TSO over VLAN requires VLAN hardware tagging. */ 3328 if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) 3329 ifp->if_capenable &= ~IFCAP_VLAN_HWTSO; 3330 reinit = 1; 3331 } 3332 if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0 && 3333 (mask & (IFCAP_HWCSUM | IFCAP_TSO4 | 3334 IFCAP_VLAN_HWTSO)) != 0) 3335 reinit = 1; 3336 if ((mask & IFCAP_WOL) != 0 && 3337 (ifp->if_capabilities & IFCAP_WOL) != 0) { 3338 if ((mask & IFCAP_WOL_UCAST) != 0) 3339 ifp->if_capenable ^= IFCAP_WOL_UCAST; 3340 if ((mask & IFCAP_WOL_MCAST) != 0) 3341 ifp->if_capenable ^= IFCAP_WOL_MCAST; 3342 if ((mask & IFCAP_WOL_MAGIC) != 0) 3343 ifp->if_capenable ^= IFCAP_WOL_MAGIC; 3344 } 3345 if (reinit && ifp->if_drv_flags & IFF_DRV_RUNNING) { 3346 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 3347 re_init(sc); 3348 } 3349 VLAN_CAPABILITIES(ifp); 3350 } 3351 break; 3352 default: 3353 error = ether_ioctl(ifp, command, data); 3354 break; 3355 } 3356 3357 return (error); 3358} 3359 3360static void 3361re_watchdog(struct rl_softc *sc) 3362{ 3363 struct ifnet *ifp; 3364 3365 RL_LOCK_ASSERT(sc); 3366 3367 if (sc->rl_watchdog_timer == 0 || --sc->rl_watchdog_timer != 0) 3368 return; 3369 3370 ifp = sc->rl_ifp; 3371 re_txeof(sc); 3372 if (sc->rl_ldata.rl_tx_free == sc->rl_ldata.rl_tx_desc_cnt) { 3373 if_printf(ifp, "watchdog timeout (missed Tx interrupts) " 3374 "-- recovering\n"); 3375 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 3376 re_start_locked(ifp); 3377 return; 3378 } 3379 3380 if_printf(ifp, "watchdog timeout\n"); 3381 ifp->if_oerrors++; 3382 3383 re_rxeof(sc, NULL); 3384 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 3385 re_init_locked(sc); 3386 if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) 3387 re_start_locked(ifp); 3388} 3389 3390/* 3391 * Stop the adapter and free any mbufs allocated to the 3392 * RX and TX lists. 3393 */ 3394static void 3395re_stop(struct rl_softc *sc) 3396{ 3397 int i; 3398 struct ifnet *ifp; 3399 struct rl_txdesc *txd; 3400 struct rl_rxdesc *rxd; 3401 3402 RL_LOCK_ASSERT(sc); 3403 3404 ifp = sc->rl_ifp; 3405 3406 sc->rl_watchdog_timer = 0; 3407 callout_stop(&sc->rl_stat_callout); 3408 ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); 3409 3410 if ((sc->rl_flags & RL_FLAG_CMDSTOP) != 0) 3411 CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_STOPREQ | RL_CMD_TX_ENB | 3412 RL_CMD_RX_ENB); 3413 else 3414 CSR_WRITE_1(sc, RL_COMMAND, 0x00); 3415 DELAY(1000); 3416 CSR_WRITE_2(sc, RL_IMR, 0x0000); 3417 CSR_WRITE_2(sc, RL_ISR, 0xFFFF); 3418 3419 if (sc->rl_head != NULL) { 3420 m_freem(sc->rl_head); 3421 sc->rl_head = sc->rl_tail = NULL; 3422 } 3423 3424 /* Free the TX list buffers. */ 3425 3426 for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) { 3427 txd = &sc->rl_ldata.rl_tx_desc[i]; 3428 if (txd->tx_m != NULL) { 3429 bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag, 3430 txd->tx_dmamap, BUS_DMASYNC_POSTWRITE); 3431 bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag, 3432 txd->tx_dmamap); 3433 m_freem(txd->tx_m); 3434 txd->tx_m = NULL; 3435 } 3436 } 3437 3438 /* Free the RX list buffers. */ 3439 3440 for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) { 3441 rxd = &sc->rl_ldata.rl_rx_desc[i]; 3442 if (rxd->rx_m != NULL) { 3443 bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag, 3444 rxd->rx_dmamap, BUS_DMASYNC_POSTREAD); 3445 bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag, 3446 rxd->rx_dmamap); 3447 m_freem(rxd->rx_m); 3448 rxd->rx_m = NULL; 3449 } 3450 } 3451} 3452 3453/* 3454 * Device suspend routine. Stop the interface and save some PCI 3455 * settings in case the BIOS doesn't restore them properly on 3456 * resume. 3457 */ 3458static int 3459re_suspend(device_t dev) 3460{ 3461 struct rl_softc *sc; 3462 3463 sc = device_get_softc(dev); 3464 3465 RL_LOCK(sc); 3466 re_stop(sc); 3467 re_setwol(sc); 3468 sc->suspended = 1; 3469 RL_UNLOCK(sc); 3470 3471 return (0); 3472} 3473 3474/* 3475 * Device resume routine. Restore some PCI settings in case the BIOS 3476 * doesn't, re-enable busmastering, and restart the interface if 3477 * appropriate. 3478 */ 3479static int 3480re_resume(device_t dev) 3481{ 3482 struct rl_softc *sc; 3483 struct ifnet *ifp; 3484 3485 sc = device_get_softc(dev); 3486 3487 RL_LOCK(sc); 3488 3489 ifp = sc->rl_ifp; 3490 /* Take controller out of sleep mode. */ 3491 if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) { 3492 if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80) 3493 CSR_WRITE_1(sc, RL_GPIO, 3494 CSR_READ_1(sc, RL_GPIO) | 0x01); 3495 } 3496 3497 /* 3498 * Clear WOL matching such that normal Rx filtering 3499 * wouldn't interfere with WOL patterns. 3500 */ 3501 re_clrwol(sc); 3502 3503 /* reinitialize interface if necessary */ 3504 if (ifp->if_flags & IFF_UP) 3505 re_init_locked(sc); 3506 3507 sc->suspended = 0; 3508 RL_UNLOCK(sc); 3509 3510 return (0); 3511} 3512 3513/* 3514 * Stop all chip I/O so that the kernel's probe routines don't 3515 * get confused by errant DMAs when rebooting. 3516 */ 3517static int 3518re_shutdown(device_t dev) 3519{ 3520 struct rl_softc *sc; 3521 3522 sc = device_get_softc(dev); 3523 3524 RL_LOCK(sc); 3525 re_stop(sc); 3526 /* 3527 * Mark interface as down since otherwise we will panic if 3528 * interrupt comes in later on, which can happen in some 3529 * cases. 3530 */ 3531 sc->rl_ifp->if_flags &= ~IFF_UP; 3532 re_setwol(sc); 3533 RL_UNLOCK(sc); 3534 3535 return (0); 3536} 3537 3538static void 3539re_setwol(struct rl_softc *sc) 3540{ 3541 struct ifnet *ifp; 3542 int pmc; 3543 uint16_t pmstat; 3544 uint8_t v; 3545 3546 RL_LOCK_ASSERT(sc); 3547 3548 if (pci_find_extcap(sc->rl_dev, PCIY_PMG, &pmc) != 0) 3549 return; 3550 3551 ifp = sc->rl_ifp; 3552 /* Put controller into sleep mode. */ 3553 if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) { 3554 if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80) 3555 CSR_WRITE_1(sc, RL_GPIO, 3556 CSR_READ_1(sc, RL_GPIO) & ~0x01); 3557 } 3558 if ((ifp->if_capenable & IFCAP_WOL) != 0 && 3559 (sc->rl_flags & RL_FLAG_WOLRXENB) != 0) 3560 CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_RX_ENB); 3561 /* Enable config register write. */ 3562 CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE); 3563 3564 /* Enable PME. */ 3565 v = CSR_READ_1(sc, RL_CFG1); 3566 v &= ~RL_CFG1_PME; 3567 if ((ifp->if_capenable & IFCAP_WOL) != 0) 3568 v |= RL_CFG1_PME; 3569 CSR_WRITE_1(sc, RL_CFG1, v); 3570 3571 v = CSR_READ_1(sc, RL_CFG3); 3572 v &= ~(RL_CFG3_WOL_LINK | RL_CFG3_WOL_MAGIC); 3573 if ((ifp->if_capenable & IFCAP_WOL_MAGIC) != 0) 3574 v |= RL_CFG3_WOL_MAGIC; 3575 CSR_WRITE_1(sc, RL_CFG3, v); 3576 3577 /* Config register write done. */ 3578 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 3579 3580 v = CSR_READ_1(sc, RL_CFG5); 3581 v &= ~(RL_CFG5_WOL_BCAST | RL_CFG5_WOL_MCAST | RL_CFG5_WOL_UCAST); 3582 v &= ~RL_CFG5_WOL_LANWAKE; 3583 if ((ifp->if_capenable & IFCAP_WOL_UCAST) != 0) 3584 v |= RL_CFG5_WOL_UCAST; 3585 if ((ifp->if_capenable & IFCAP_WOL_MCAST) != 0) 3586 v |= RL_CFG5_WOL_MCAST | RL_CFG5_WOL_BCAST; 3587 if ((ifp->if_capenable & IFCAP_WOL) != 0) 3588 v |= RL_CFG5_WOL_LANWAKE; 3589 CSR_WRITE_1(sc, RL_CFG5, v); 3590 3591 if ((ifp->if_capenable & IFCAP_WOL) != 0 && 3592 (sc->rl_flags & RL_FLAG_PHYWAKE_PM) != 0) 3593 CSR_WRITE_1(sc, RL_PMCH, CSR_READ_1(sc, RL_PMCH) & ~0x80); 3594 /* 3595 * It seems that hardware resets its link speed to 100Mbps in 3596 * power down mode so switching to 100Mbps in driver is not 3597 * needed. 3598 */ 3599 3600 /* Request PME if WOL is requested. */ 3601 pmstat = pci_read_config(sc->rl_dev, pmc + PCIR_POWER_STATUS, 2); 3602 pmstat &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE); 3603 if ((ifp->if_capenable & IFCAP_WOL) != 0) 3604 pmstat |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE; 3605 pci_write_config(sc->rl_dev, pmc + PCIR_POWER_STATUS, pmstat, 2); 3606} 3607 3608static void 3609re_clrwol(struct rl_softc *sc) 3610{ 3611 int pmc; 3612 uint8_t v; 3613 3614 RL_LOCK_ASSERT(sc); 3615 3616 if (pci_find_extcap(sc->rl_dev, PCIY_PMG, &pmc) != 0) 3617 return; 3618 3619 /* Enable config register write. */ 3620 CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE); 3621 3622 v = CSR_READ_1(sc, RL_CFG3); 3623 v &= ~(RL_CFG3_WOL_LINK | RL_CFG3_WOL_MAGIC); 3624 CSR_WRITE_1(sc, RL_CFG3, v); 3625 3626 /* Config register write done. */ 3627 CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF); 3628 3629 v = CSR_READ_1(sc, RL_CFG5); 3630 v &= ~(RL_CFG5_WOL_BCAST | RL_CFG5_WOL_MCAST | RL_CFG5_WOL_UCAST); 3631 v &= ~RL_CFG5_WOL_LANWAKE; 3632 CSR_WRITE_1(sc, RL_CFG5, v); 3633} 3634 3635static void 3636re_add_sysctls(struct rl_softc *sc) 3637{ 3638 struct sysctl_ctx_list *ctx; 3639 struct sysctl_oid_list *children; 3640 int error; 3641 3642 ctx = device_get_sysctl_ctx(sc->rl_dev); 3643 children = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->rl_dev)); 3644 3645 SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "stats", 3646 CTLTYPE_INT | CTLFLAG_RW, sc, 0, re_sysctl_stats, "I", 3647 "Statistics Information"); 3648 if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) == 0) 3649 return; 3650 3651 SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "int_rx_mod", 3652 CTLTYPE_INT | CTLFLAG_RW, &sc->rl_int_rx_mod, 0, 3653 sysctl_hw_re_int_mod, "I", "re RX interrupt moderation"); 3654 /* Pull in device tunables. */ 3655 sc->rl_int_rx_mod = RL_TIMER_DEFAULT; 3656 error = resource_int_value(device_get_name(sc->rl_dev), 3657 device_get_unit(sc->rl_dev), "int_rx_mod", &sc->rl_int_rx_mod); 3658 if (error == 0) { 3659 if (sc->rl_int_rx_mod < RL_TIMER_MIN || 3660 sc->rl_int_rx_mod > RL_TIMER_MAX) { 3661 device_printf(sc->rl_dev, "int_rx_mod value out of " 3662 "range; using default: %d\n", 3663 RL_TIMER_DEFAULT); 3664 sc->rl_int_rx_mod = RL_TIMER_DEFAULT; 3665 } 3666 } 3667 3668} 3669 3670static int 3671re_sysctl_stats(SYSCTL_HANDLER_ARGS) 3672{ 3673 struct rl_softc *sc; 3674 struct rl_stats *stats; 3675 int error, i, result; 3676 3677 result = -1; 3678 error = sysctl_handle_int(oidp, &result, 0, req); 3679 if (error || req->newptr == NULL) 3680 return (error); 3681 3682 if (result == 1) { 3683 sc = (struct rl_softc *)arg1; 3684 RL_LOCK(sc); 3685 if ((sc->rl_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 3686 RL_UNLOCK(sc); 3687 goto done; 3688 } 3689 bus_dmamap_sync(sc->rl_ldata.rl_stag, 3690 sc->rl_ldata.rl_smap, BUS_DMASYNC_PREREAD); 3691 CSR_WRITE_4(sc, RL_DUMPSTATS_HI, 3692 RL_ADDR_HI(sc->rl_ldata.rl_stats_addr)); 3693 CSR_WRITE_4(sc, RL_DUMPSTATS_LO, 3694 RL_ADDR_LO(sc->rl_ldata.rl_stats_addr)); 3695 CSR_WRITE_4(sc, RL_DUMPSTATS_LO, 3696 RL_ADDR_LO(sc->rl_ldata.rl_stats_addr | 3697 RL_DUMPSTATS_START)); 3698 for (i = RL_TIMEOUT; i > 0; i--) { 3699 if ((CSR_READ_4(sc, RL_DUMPSTATS_LO) & 3700 RL_DUMPSTATS_START) == 0) 3701 break; 3702 DELAY(1000); 3703 } 3704 bus_dmamap_sync(sc->rl_ldata.rl_stag, 3705 sc->rl_ldata.rl_smap, BUS_DMASYNC_POSTREAD); 3706 RL_UNLOCK(sc); 3707 if (i == 0) { 3708 device_printf(sc->rl_dev, 3709 "DUMP statistics request timedout\n"); 3710 return (ETIMEDOUT); 3711 } 3712done: 3713 stats = sc->rl_ldata.rl_stats; 3714 printf("%s statistics:\n", device_get_nameunit(sc->rl_dev)); 3715 printf("Tx frames : %ju\n", 3716 (uintmax_t)le64toh(stats->rl_tx_pkts)); 3717 printf("Rx frames : %ju\n", 3718 (uintmax_t)le64toh(stats->rl_rx_pkts)); 3719 printf("Tx errors : %ju\n", 3720 (uintmax_t)le64toh(stats->rl_tx_errs)); 3721 printf("Rx errors : %u\n", 3722 le32toh(stats->rl_rx_errs)); 3723 printf("Rx missed frames : %u\n", 3724 (uint32_t)le16toh(stats->rl_missed_pkts)); 3725 printf("Rx frame alignment errs : %u\n", 3726 (uint32_t)le16toh(stats->rl_rx_framealign_errs)); 3727 printf("Tx single collisions : %u\n", 3728 le32toh(stats->rl_tx_onecoll)); 3729 printf("Tx multiple collisions : %u\n", 3730 le32toh(stats->rl_tx_multicolls)); 3731 printf("Rx unicast frames : %ju\n", 3732 (uintmax_t)le64toh(stats->rl_rx_ucasts)); 3733 printf("Rx broadcast frames : %ju\n", 3734 (uintmax_t)le64toh(stats->rl_rx_bcasts)); 3735 printf("Rx multicast frames : %u\n", 3736 le32toh(stats->rl_rx_mcasts)); 3737 printf("Tx aborts : %u\n", 3738 (uint32_t)le16toh(stats->rl_tx_aborts)); 3739 printf("Tx underruns : %u\n", 3740 (uint32_t)le16toh(stats->rl_rx_underruns)); 3741 } 3742 3743 return (error); 3744} 3745 3746static int 3747sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high) 3748{ 3749 int error, value; 3750 3751 if (arg1 == NULL) 3752 return (EINVAL); 3753 value = *(int *)arg1; 3754 error = sysctl_handle_int(oidp, &value, 0, req); 3755 if (error || req->newptr == NULL) 3756 return (error); 3757 if (value < low || value > high) 3758 return (EINVAL); 3759 *(int *)arg1 = value; 3760 3761 return (0); 3762} 3763 3764static int 3765sysctl_hw_re_int_mod(SYSCTL_HANDLER_ARGS) 3766{ 3767 3768 return (sysctl_int_range(oidp, arg1, arg2, req, RL_TIMER_MIN, 3769 RL_TIMER_MAX)); 3770} 3771