xref: /freebsd/sys/dev/re/if_re.c (revision b3aaa0cc21c63d388230c7ef2a80abd631ff20d5)
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("$FreeBSD$");
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/taskqueue.h>
127 
128 #include <net/if.h>
129 #include <net/if_arp.h>
130 #include <net/ethernet.h>
131 #include <net/if_dl.h>
132 #include <net/if_media.h>
133 #include <net/if_types.h>
134 #include <net/if_vlan_var.h>
135 
136 #include <net/bpf.h>
137 
138 #include <machine/bus.h>
139 #include <machine/resource.h>
140 #include <sys/bus.h>
141 #include <sys/rman.h>
142 
143 #include <dev/mii/mii.h>
144 #include <dev/mii/miivar.h>
145 
146 #include <dev/pci/pcireg.h>
147 #include <dev/pci/pcivar.h>
148 
149 #include <pci/if_rlreg.h>
150 
151 MODULE_DEPEND(re, pci, 1, 1, 1);
152 MODULE_DEPEND(re, ether, 1, 1, 1);
153 MODULE_DEPEND(re, miibus, 1, 1, 1);
154 
155 /* "device miibus" required.  See GENERIC if you get errors here. */
156 #include "miibus_if.h"
157 
158 /* Tunables. */
159 static int msi_disable = 0;
160 TUNABLE_INT("hw.re.msi_disable", &msi_disable);
161 
162 #define RE_CSUM_FEATURES    (CSUM_IP | CSUM_TCP | CSUM_UDP)
163 
164 /*
165  * Various supported device vendors/types and their names.
166  */
167 static struct rl_type re_devs[] = {
168 	{ DLINK_VENDORID, DLINK_DEVICEID_528T, 0,
169 	    "D-Link DGE-528(T) Gigabit Ethernet Adapter" },
170 	{ RT_VENDORID, RT_DEVICEID_8139, 0,
171 	    "RealTek 8139C+ 10/100BaseTX" },
172 	{ RT_VENDORID, RT_DEVICEID_8101E, 0,
173 	    "RealTek 8101E/8102E/8102EL PCIe 10/100baseTX" },
174 	{ RT_VENDORID, RT_DEVICEID_8168, 0,
175 	    "RealTek 8168/8168B/8168C/8168CP/8168D/8111B/8111C/8111CP PCIe "
176 	    "Gigabit Ethernet" },
177 	{ RT_VENDORID, RT_DEVICEID_8169, 0,
178 	    "RealTek 8169/8169S/8169SB(L)/8110S/8110SB(L) Gigabit Ethernet" },
179 	{ RT_VENDORID, RT_DEVICEID_8169SC, 0,
180 	    "RealTek 8169SC/8110SC Single-chip Gigabit Ethernet" },
181 	{ COREGA_VENDORID, COREGA_DEVICEID_CGLAPCIGT, 0,
182 	    "Corega CG-LAPCIGT (RTL8169S) Gigabit Ethernet" },
183 	{ LINKSYS_VENDORID, LINKSYS_DEVICEID_EG1032, 0,
184 	    "Linksys EG1032 (RTL8169S) Gigabit Ethernet" },
185 	{ USR_VENDORID, USR_DEVICEID_997902, 0,
186 	    "US Robotics 997902 (RTL8169S) Gigabit Ethernet" }
187 };
188 
189 static struct rl_hwrev re_hwrevs[] = {
190 	{ RL_HWREV_8139, RL_8139,  "" },
191 	{ RL_HWREV_8139A, RL_8139, "A" },
192 	{ RL_HWREV_8139AG, RL_8139, "A-G" },
193 	{ RL_HWREV_8139B, RL_8139, "B" },
194 	{ RL_HWREV_8130, RL_8139, "8130" },
195 	{ RL_HWREV_8139C, RL_8139, "C" },
196 	{ RL_HWREV_8139D, RL_8139, "8139D/8100B/8100C" },
197 	{ RL_HWREV_8139CPLUS, RL_8139CPLUS, "C+"},
198 	{ RL_HWREV_8168_SPIN1, RL_8169, "8168"},
199 	{ RL_HWREV_8169, RL_8169, "8169"},
200 	{ RL_HWREV_8169S, RL_8169, "8169S"},
201 	{ RL_HWREV_8110S, RL_8169, "8110S"},
202 	{ RL_HWREV_8169_8110SB, RL_8169, "8169SB/8110SB"},
203 	{ RL_HWREV_8169_8110SC, RL_8169, "8169SC/8110SC"},
204 	{ RL_HWREV_8169_8110SBL, RL_8169, "8169SBL/8110SBL"},
205 	{ RL_HWREV_8169_8110SCE, RL_8169, "8169SC/8110SC"},
206 	{ RL_HWREV_8100, RL_8139, "8100"},
207 	{ RL_HWREV_8101, RL_8139, "8101"},
208 	{ RL_HWREV_8100E, RL_8169, "8100E"},
209 	{ RL_HWREV_8101E, RL_8169, "8101E"},
210 	{ RL_HWREV_8102E, RL_8169, "8102E"},
211 	{ RL_HWREV_8102EL, RL_8169, "8102EL"},
212 	{ RL_HWREV_8168_SPIN2, RL_8169, "8168"},
213 	{ RL_HWREV_8168_SPIN3, RL_8169, "8168"},
214 	{ RL_HWREV_8168C, RL_8169, "8168C/8111C"},
215 	{ RL_HWREV_8168C_SPIN2, RL_8169, "8168C/8111C"},
216 	{ RL_HWREV_8168CP, RL_8169, "8168CP/8111CP"},
217 	{ RL_HWREV_8168D, RL_8169, "8168D"},
218 	{ 0, 0, NULL }
219 };
220 
221 static int re_probe		(device_t);
222 static int re_attach		(device_t);
223 static int re_detach		(device_t);
224 
225 static int re_encap		(struct rl_softc *, struct mbuf **);
226 
227 static void re_dma_map_addr	(void *, bus_dma_segment_t *, int, int);
228 static int re_allocmem		(device_t, struct rl_softc *);
229 static __inline void re_discard_rxbuf
230 				(struct rl_softc *, int);
231 static int re_newbuf		(struct rl_softc *, int);
232 static int re_rx_list_init	(struct rl_softc *);
233 static int re_tx_list_init	(struct rl_softc *);
234 #ifdef RE_FIXUP_RX
235 static __inline void re_fixup_rx
236 				(struct mbuf *);
237 #endif
238 static int re_rxeof		(struct rl_softc *);
239 static void re_txeof		(struct rl_softc *);
240 #ifdef DEVICE_POLLING
241 static void re_poll		(struct ifnet *, enum poll_cmd, int);
242 static void re_poll_locked	(struct ifnet *, enum poll_cmd, int);
243 #endif
244 static int re_intr		(void *);
245 static void re_tick		(void *);
246 static void re_tx_task		(void *, int);
247 static void re_int_task		(void *, int);
248 static void re_start		(struct ifnet *);
249 static int re_ioctl		(struct ifnet *, u_long, caddr_t);
250 static void re_init		(void *);
251 static void re_init_locked	(struct rl_softc *);
252 static void re_stop		(struct rl_softc *);
253 static void re_watchdog		(struct rl_softc *);
254 static int re_suspend		(device_t);
255 static int re_resume		(device_t);
256 static int re_shutdown		(device_t);
257 static int re_ifmedia_upd	(struct ifnet *);
258 static void re_ifmedia_sts	(struct ifnet *, struct ifmediareq *);
259 
260 static void re_eeprom_putbyte	(struct rl_softc *, int);
261 static void re_eeprom_getword	(struct rl_softc *, int, u_int16_t *);
262 static void re_read_eeprom	(struct rl_softc *, caddr_t, int, int);
263 static int re_gmii_readreg	(device_t, int, int);
264 static int re_gmii_writereg	(device_t, int, int, int);
265 
266 static int re_miibus_readreg	(device_t, int, int);
267 static int re_miibus_writereg	(device_t, int, int, int);
268 static void re_miibus_statchg	(device_t);
269 
270 static void re_set_rxmode		(struct rl_softc *);
271 static void re_reset		(struct rl_softc *);
272 static void re_setwol		(struct rl_softc *);
273 static void re_clrwol		(struct rl_softc *);
274 
275 #ifdef RE_DIAG
276 static int re_diag		(struct rl_softc *);
277 #endif
278 
279 static device_method_t re_methods[] = {
280 	/* Device interface */
281 	DEVMETHOD(device_probe,		re_probe),
282 	DEVMETHOD(device_attach,	re_attach),
283 	DEVMETHOD(device_detach,	re_detach),
284 	DEVMETHOD(device_suspend,	re_suspend),
285 	DEVMETHOD(device_resume,	re_resume),
286 	DEVMETHOD(device_shutdown,	re_shutdown),
287 
288 	/* bus interface */
289 	DEVMETHOD(bus_print_child,	bus_generic_print_child),
290 	DEVMETHOD(bus_driver_added,	bus_generic_driver_added),
291 
292 	/* MII interface */
293 	DEVMETHOD(miibus_readreg,	re_miibus_readreg),
294 	DEVMETHOD(miibus_writereg,	re_miibus_writereg),
295 	DEVMETHOD(miibus_statchg,	re_miibus_statchg),
296 
297 	{ 0, 0 }
298 };
299 
300 static driver_t re_driver = {
301 	"re",
302 	re_methods,
303 	sizeof(struct rl_softc)
304 };
305 
306 static devclass_t re_devclass;
307 
308 DRIVER_MODULE(re, pci, re_driver, re_devclass, 0, 0);
309 DRIVER_MODULE(re, cardbus, re_driver, re_devclass, 0, 0);
310 DRIVER_MODULE(miibus, re, miibus_driver, miibus_devclass, 0, 0);
311 
312 #define EE_SET(x)					\
313 	CSR_WRITE_1(sc, RL_EECMD,			\
314 		CSR_READ_1(sc, RL_EECMD) | x)
315 
316 #define EE_CLR(x)					\
317 	CSR_WRITE_1(sc, RL_EECMD,			\
318 		CSR_READ_1(sc, RL_EECMD) & ~x)
319 
320 /*
321  * Send a read command and address to the EEPROM, check for ACK.
322  */
323 static void
324 re_eeprom_putbyte(struct rl_softc *sc, int addr)
325 {
326 	int			d, i;
327 
328 	d = addr | (RL_9346_READ << sc->rl_eewidth);
329 
330 	/*
331 	 * Feed in each bit and strobe the clock.
332 	 */
333 
334 	for (i = 1 << (sc->rl_eewidth + 3); i; i >>= 1) {
335 		if (d & i) {
336 			EE_SET(RL_EE_DATAIN);
337 		} else {
338 			EE_CLR(RL_EE_DATAIN);
339 		}
340 		DELAY(100);
341 		EE_SET(RL_EE_CLK);
342 		DELAY(150);
343 		EE_CLR(RL_EE_CLK);
344 		DELAY(100);
345 	}
346 }
347 
348 /*
349  * Read a word of data stored in the EEPROM at address 'addr.'
350  */
351 static void
352 re_eeprom_getword(struct rl_softc *sc, int addr, u_int16_t *dest)
353 {
354 	int			i;
355 	u_int16_t		word = 0;
356 
357 	/*
358 	 * Send address of word we want to read.
359 	 */
360 	re_eeprom_putbyte(sc, addr);
361 
362 	/*
363 	 * Start reading bits from EEPROM.
364 	 */
365 	for (i = 0x8000; i; i >>= 1) {
366 		EE_SET(RL_EE_CLK);
367 		DELAY(100);
368 		if (CSR_READ_1(sc, RL_EECMD) & RL_EE_DATAOUT)
369 			word |= i;
370 		EE_CLR(RL_EE_CLK);
371 		DELAY(100);
372 	}
373 
374 	*dest = word;
375 }
376 
377 /*
378  * Read a sequence of words from the EEPROM.
379  */
380 static void
381 re_read_eeprom(struct rl_softc *sc, caddr_t dest, int off, int cnt)
382 {
383 	int			i;
384 	u_int16_t		word = 0, *ptr;
385 
386 	CSR_SETBIT_1(sc, RL_EECMD, RL_EEMODE_PROGRAM);
387 
388         DELAY(100);
389 
390 	for (i = 0; i < cnt; i++) {
391 		CSR_SETBIT_1(sc, RL_EECMD, RL_EE_SEL);
392 		re_eeprom_getword(sc, off + i, &word);
393 		CSR_CLRBIT_1(sc, RL_EECMD, RL_EE_SEL);
394 		ptr = (u_int16_t *)(dest + (i * 2));
395                 *ptr = word;
396 	}
397 
398 	CSR_CLRBIT_1(sc, RL_EECMD, RL_EEMODE_PROGRAM);
399 }
400 
401 static int
402 re_gmii_readreg(device_t dev, int phy, int reg)
403 {
404 	struct rl_softc		*sc;
405 	u_int32_t		rval;
406 	int			i;
407 
408 	if (phy != 1)
409 		return (0);
410 
411 	sc = device_get_softc(dev);
412 
413 	/* Let the rgephy driver read the GMEDIASTAT register */
414 
415 	if (reg == RL_GMEDIASTAT) {
416 		rval = CSR_READ_1(sc, RL_GMEDIASTAT);
417 		return (rval);
418 	}
419 
420 	CSR_WRITE_4(sc, RL_PHYAR, reg << 16);
421 
422 	for (i = 0; i < RL_PHY_TIMEOUT; i++) {
423 		rval = CSR_READ_4(sc, RL_PHYAR);
424 		if (rval & RL_PHYAR_BUSY)
425 			break;
426 		DELAY(100);
427 	}
428 
429 	if (i == RL_PHY_TIMEOUT) {
430 		device_printf(sc->rl_dev, "PHY read failed\n");
431 		return (0);
432 	}
433 
434 	return (rval & RL_PHYAR_PHYDATA);
435 }
436 
437 static int
438 re_gmii_writereg(device_t dev, int phy, int reg, int data)
439 {
440 	struct rl_softc		*sc;
441 	u_int32_t		rval;
442 	int			i;
443 
444 	sc = device_get_softc(dev);
445 
446 	CSR_WRITE_4(sc, RL_PHYAR, (reg << 16) |
447 	    (data & RL_PHYAR_PHYDATA) | RL_PHYAR_BUSY);
448 
449 	for (i = 0; i < RL_PHY_TIMEOUT; i++) {
450 		rval = CSR_READ_4(sc, RL_PHYAR);
451 		if (!(rval & RL_PHYAR_BUSY))
452 			break;
453 		DELAY(100);
454 	}
455 
456 	if (i == RL_PHY_TIMEOUT) {
457 		device_printf(sc->rl_dev, "PHY write failed\n");
458 		return (0);
459 	}
460 
461 	return (0);
462 }
463 
464 static int
465 re_miibus_readreg(device_t dev, int phy, int reg)
466 {
467 	struct rl_softc		*sc;
468 	u_int16_t		rval = 0;
469 	u_int16_t		re8139_reg = 0;
470 
471 	sc = device_get_softc(dev);
472 
473 	if (sc->rl_type == RL_8169) {
474 		rval = re_gmii_readreg(dev, phy, reg);
475 		return (rval);
476 	}
477 
478 	/* Pretend the internal PHY is only at address 0 */
479 	if (phy) {
480 		return (0);
481 	}
482 	switch (reg) {
483 	case MII_BMCR:
484 		re8139_reg = RL_BMCR;
485 		break;
486 	case MII_BMSR:
487 		re8139_reg = RL_BMSR;
488 		break;
489 	case MII_ANAR:
490 		re8139_reg = RL_ANAR;
491 		break;
492 	case MII_ANER:
493 		re8139_reg = RL_ANER;
494 		break;
495 	case MII_ANLPAR:
496 		re8139_reg = RL_LPAR;
497 		break;
498 	case MII_PHYIDR1:
499 	case MII_PHYIDR2:
500 		return (0);
501 	/*
502 	 * Allow the rlphy driver to read the media status
503 	 * register. If we have a link partner which does not
504 	 * support NWAY, this is the register which will tell
505 	 * us the results of parallel detection.
506 	 */
507 	case RL_MEDIASTAT:
508 		rval = CSR_READ_1(sc, RL_MEDIASTAT);
509 		return (rval);
510 	default:
511 		device_printf(sc->rl_dev, "bad phy register\n");
512 		return (0);
513 	}
514 	rval = CSR_READ_2(sc, re8139_reg);
515 	if (sc->rl_type == RL_8139CPLUS && re8139_reg == RL_BMCR) {
516 		/* 8139C+ has different bit layout. */
517 		rval &= ~(BMCR_LOOP | BMCR_ISO);
518 	}
519 	return (rval);
520 }
521 
522 static int
523 re_miibus_writereg(device_t dev, int phy, int reg, int data)
524 {
525 	struct rl_softc		*sc;
526 	u_int16_t		re8139_reg = 0;
527 	int			rval = 0;
528 
529 	sc = device_get_softc(dev);
530 
531 	if (sc->rl_type == RL_8169) {
532 		rval = re_gmii_writereg(dev, phy, reg, data);
533 		return (rval);
534 	}
535 
536 	/* Pretend the internal PHY is only at address 0 */
537 	if (phy)
538 		return (0);
539 
540 	switch (reg) {
541 	case MII_BMCR:
542 		re8139_reg = RL_BMCR;
543 		if (sc->rl_type == RL_8139CPLUS) {
544 			/* 8139C+ has different bit layout. */
545 			data &= ~(BMCR_LOOP | BMCR_ISO);
546 		}
547 		break;
548 	case MII_BMSR:
549 		re8139_reg = RL_BMSR;
550 		break;
551 	case MII_ANAR:
552 		re8139_reg = RL_ANAR;
553 		break;
554 	case MII_ANER:
555 		re8139_reg = RL_ANER;
556 		break;
557 	case MII_ANLPAR:
558 		re8139_reg = RL_LPAR;
559 		break;
560 	case MII_PHYIDR1:
561 	case MII_PHYIDR2:
562 		return (0);
563 		break;
564 	default:
565 		device_printf(sc->rl_dev, "bad phy register\n");
566 		return (0);
567 	}
568 	CSR_WRITE_2(sc, re8139_reg, data);
569 	return (0);
570 }
571 
572 static void
573 re_miibus_statchg(device_t dev)
574 {
575 	struct rl_softc		*sc;
576 	struct ifnet		*ifp;
577 	struct mii_data		*mii;
578 
579 	sc = device_get_softc(dev);
580 	mii = device_get_softc(sc->rl_miibus);
581 	ifp = sc->rl_ifp;
582 	if (mii == NULL || ifp == NULL ||
583 	    (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
584 		return;
585 
586 	sc->rl_flags &= ~RL_FLAG_LINK;
587 	if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
588 	    (IFM_ACTIVE | IFM_AVALID)) {
589 		switch (IFM_SUBTYPE(mii->mii_media_active)) {
590 		case IFM_10_T:
591 		case IFM_100_TX:
592 			sc->rl_flags |= RL_FLAG_LINK;
593 			break;
594 		case IFM_1000_T:
595 			if ((sc->rl_flags & RL_FLAG_FASTETHER) != 0)
596 				break;
597 			sc->rl_flags |= RL_FLAG_LINK;
598 			break;
599 		default:
600 			break;
601 		}
602 	}
603 	/*
604 	 * RealTek controllers does not provide any interface to
605 	 * Tx/Rx MACs for resolved speed, duplex and flow-control
606 	 * parameters.
607 	 */
608 }
609 
610 /*
611  * Set the RX configuration and 64-bit multicast hash filter.
612  */
613 static void
614 re_set_rxmode(struct rl_softc *sc)
615 {
616 	struct ifnet		*ifp;
617 	struct ifmultiaddr	*ifma;
618 	uint32_t		hashes[2] = { 0, 0 };
619 	uint32_t		h, rxfilt;
620 
621 	RL_LOCK_ASSERT(sc);
622 
623 	ifp = sc->rl_ifp;
624 
625 	rxfilt = RL_RXCFG_CONFIG | RL_RXCFG_RX_INDIV | RL_RXCFG_RX_BROAD;
626 
627 	if (ifp->if_flags & (IFF_ALLMULTI | IFF_PROMISC)) {
628 		if (ifp->if_flags & IFF_PROMISC)
629 			rxfilt |= RL_RXCFG_RX_ALLPHYS;
630 		/*
631 		 * Unlike other hardwares, we have to explicitly set
632 		 * RL_RXCFG_RX_MULTI to receive multicast frames in
633 		 * promiscuous mode.
634 		 */
635 		rxfilt |= RL_RXCFG_RX_MULTI;
636 		hashes[0] = hashes[1] = 0xffffffff;
637 		goto done;
638 	}
639 
640 	IF_ADDR_LOCK(ifp);
641 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
642 		if (ifma->ifma_addr->sa_family != AF_LINK)
643 			continue;
644 		h = ether_crc32_be(LLADDR((struct sockaddr_dl *)
645 		    ifma->ifma_addr), ETHER_ADDR_LEN) >> 26;
646 		if (h < 32)
647 			hashes[0] |= (1 << h);
648 		else
649 			hashes[1] |= (1 << (h - 32));
650 	}
651 	IF_ADDR_UNLOCK(ifp);
652 
653 	if (hashes[0] != 0 || hashes[1] != 0) {
654 		/*
655 		 * For some unfathomable reason, RealTek decided to
656 		 * reverse the order of the multicast hash registers
657 		 * in the PCI Express parts.  This means we have to
658 		 * write the hash pattern in reverse order for those
659 		 * devices.
660 		 */
661 		if ((sc->rl_flags & RL_FLAG_PCIE) != 0) {
662 			h = bswap32(hashes[0]);
663 			hashes[0] = bswap32(hashes[1]);
664 			hashes[1] = h;
665 		}
666 		rxfilt |= RL_RXCFG_RX_MULTI;
667 	}
668 
669 done:
670 	CSR_WRITE_4(sc, RL_MAR0, hashes[0]);
671 	CSR_WRITE_4(sc, RL_MAR4, hashes[1]);
672 	CSR_WRITE_4(sc, RL_RXCFG, rxfilt);
673 }
674 
675 static void
676 re_reset(struct rl_softc *sc)
677 {
678 	int			i;
679 
680 	RL_LOCK_ASSERT(sc);
681 
682 	CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_RESET);
683 
684 	for (i = 0; i < RL_TIMEOUT; i++) {
685 		DELAY(10);
686 		if (!(CSR_READ_1(sc, RL_COMMAND) & RL_CMD_RESET))
687 			break;
688 	}
689 	if (i == RL_TIMEOUT)
690 		device_printf(sc->rl_dev, "reset never completed!\n");
691 
692 	if ((sc->rl_flags & RL_FLAG_MACRESET) != 0)
693 		CSR_WRITE_1(sc, 0x82, 1);
694 	if (sc->rl_hwrev == RL_HWREV_8169S)
695 		re_gmii_writereg(sc->rl_dev, 1, 0x0b, 0);
696 }
697 
698 #ifdef RE_DIAG
699 
700 /*
701  * The following routine is designed to test for a defect on some
702  * 32-bit 8169 cards. Some of these NICs have the REQ64# and ACK64#
703  * lines connected to the bus, however for a 32-bit only card, they
704  * should be pulled high. The result of this defect is that the
705  * NIC will not work right if you plug it into a 64-bit slot: DMA
706  * operations will be done with 64-bit transfers, which will fail
707  * because the 64-bit data lines aren't connected.
708  *
709  * There's no way to work around this (short of talking a soldering
710  * iron to the board), however we can detect it. The method we use
711  * here is to put the NIC into digital loopback mode, set the receiver
712  * to promiscuous mode, and then try to send a frame. We then compare
713  * the frame data we sent to what was received. If the data matches,
714  * then the NIC is working correctly, otherwise we know the user has
715  * a defective NIC which has been mistakenly plugged into a 64-bit PCI
716  * slot. In the latter case, there's no way the NIC can work correctly,
717  * so we print out a message on the console and abort the device attach.
718  */
719 
720 static int
721 re_diag(struct rl_softc *sc)
722 {
723 	struct ifnet		*ifp = sc->rl_ifp;
724 	struct mbuf		*m0;
725 	struct ether_header	*eh;
726 	struct rl_desc		*cur_rx;
727 	u_int16_t		status;
728 	u_int32_t		rxstat;
729 	int			total_len, i, error = 0, phyaddr;
730 	u_int8_t		dst[] = { 0x00, 'h', 'e', 'l', 'l', 'o' };
731 	u_int8_t		src[] = { 0x00, 'w', 'o', 'r', 'l', 'd' };
732 
733 	/* Allocate a single mbuf */
734 	MGETHDR(m0, M_DONTWAIT, MT_DATA);
735 	if (m0 == NULL)
736 		return (ENOBUFS);
737 
738 	RL_LOCK(sc);
739 
740 	/*
741 	 * Initialize the NIC in test mode. This sets the chip up
742 	 * so that it can send and receive frames, but performs the
743 	 * following special functions:
744 	 * - Puts receiver in promiscuous mode
745 	 * - Enables digital loopback mode
746 	 * - Leaves interrupts turned off
747 	 */
748 
749 	ifp->if_flags |= IFF_PROMISC;
750 	sc->rl_testmode = 1;
751 	re_init_locked(sc);
752 	sc->rl_flags |= RL_FLAG_LINK;
753 	if (sc->rl_type == RL_8169)
754 		phyaddr = 1;
755 	else
756 		phyaddr = 0;
757 
758 	re_miibus_writereg(sc->rl_dev, phyaddr, MII_BMCR, BMCR_RESET);
759 	for (i = 0; i < RL_TIMEOUT; i++) {
760 		status = re_miibus_readreg(sc->rl_dev, phyaddr, MII_BMCR);
761 		if (!(status & BMCR_RESET))
762 			break;
763 	}
764 
765 	re_miibus_writereg(sc->rl_dev, phyaddr, MII_BMCR, BMCR_LOOP);
766 	CSR_WRITE_2(sc, RL_ISR, RL_INTRS);
767 
768 	DELAY(100000);
769 
770 	/* Put some data in the mbuf */
771 
772 	eh = mtod(m0, struct ether_header *);
773 	bcopy ((char *)&dst, eh->ether_dhost, ETHER_ADDR_LEN);
774 	bcopy ((char *)&src, eh->ether_shost, ETHER_ADDR_LEN);
775 	eh->ether_type = htons(ETHERTYPE_IP);
776 	m0->m_pkthdr.len = m0->m_len = ETHER_MIN_LEN - ETHER_CRC_LEN;
777 
778 	/*
779 	 * Queue the packet, start transmission.
780 	 * Note: IF_HANDOFF() ultimately calls re_start() for us.
781 	 */
782 
783 	CSR_WRITE_2(sc, RL_ISR, 0xFFFF);
784 	RL_UNLOCK(sc);
785 	/* XXX: re_diag must not be called when in ALTQ mode */
786 	IF_HANDOFF(&ifp->if_snd, m0, ifp);
787 	RL_LOCK(sc);
788 	m0 = NULL;
789 
790 	/* Wait for it to propagate through the chip */
791 
792 	DELAY(100000);
793 	for (i = 0; i < RL_TIMEOUT; i++) {
794 		status = CSR_READ_2(sc, RL_ISR);
795 		CSR_WRITE_2(sc, RL_ISR, status);
796 		if ((status & (RL_ISR_TIMEOUT_EXPIRED|RL_ISR_RX_OK)) ==
797 		    (RL_ISR_TIMEOUT_EXPIRED|RL_ISR_RX_OK))
798 			break;
799 		DELAY(10);
800 	}
801 
802 	if (i == RL_TIMEOUT) {
803 		device_printf(sc->rl_dev,
804 		    "diagnostic failed, failed to receive packet in"
805 		    " loopback mode\n");
806 		error = EIO;
807 		goto done;
808 	}
809 
810 	/*
811 	 * The packet should have been dumped into the first
812 	 * entry in the RX DMA ring. Grab it from there.
813 	 */
814 
815 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
816 	    sc->rl_ldata.rl_rx_list_map,
817 	    BUS_DMASYNC_POSTREAD);
818 	bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag,
819 	    sc->rl_ldata.rl_rx_desc[0].rx_dmamap,
820 	    BUS_DMASYNC_POSTREAD);
821 	bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag,
822 	    sc->rl_ldata.rl_rx_desc[0].rx_dmamap);
823 
824 	m0 = sc->rl_ldata.rl_rx_desc[0].rx_m;
825 	sc->rl_ldata.rl_rx_desc[0].rx_m = NULL;
826 	eh = mtod(m0, struct ether_header *);
827 
828 	cur_rx = &sc->rl_ldata.rl_rx_list[0];
829 	total_len = RL_RXBYTES(cur_rx);
830 	rxstat = le32toh(cur_rx->rl_cmdstat);
831 
832 	if (total_len != ETHER_MIN_LEN) {
833 		device_printf(sc->rl_dev,
834 		    "diagnostic failed, received short packet\n");
835 		error = EIO;
836 		goto done;
837 	}
838 
839 	/* Test that the received packet data matches what we sent. */
840 
841 	if (bcmp((char *)&eh->ether_dhost, (char *)&dst, ETHER_ADDR_LEN) ||
842 	    bcmp((char *)&eh->ether_shost, (char *)&src, ETHER_ADDR_LEN) ||
843 	    ntohs(eh->ether_type) != ETHERTYPE_IP) {
844 		device_printf(sc->rl_dev, "WARNING, DMA FAILURE!\n");
845 		device_printf(sc->rl_dev, "expected TX data: %6D/%6D/0x%x\n",
846 		    dst, ":", src, ":", ETHERTYPE_IP);
847 		device_printf(sc->rl_dev, "received RX data: %6D/%6D/0x%x\n",
848 		    eh->ether_dhost, ":",  eh->ether_shost, ":",
849 		    ntohs(eh->ether_type));
850 		device_printf(sc->rl_dev, "You may have a defective 32-bit "
851 		    "NIC plugged into a 64-bit PCI slot.\n");
852 		device_printf(sc->rl_dev, "Please re-install the NIC in a "
853 		    "32-bit slot for proper operation.\n");
854 		device_printf(sc->rl_dev, "Read the re(4) man page for more "
855 		    "details.\n");
856 		error = EIO;
857 	}
858 
859 done:
860 	/* Turn interface off, release resources */
861 
862 	sc->rl_testmode = 0;
863 	sc->rl_flags &= ~RL_FLAG_LINK;
864 	ifp->if_flags &= ~IFF_PROMISC;
865 	re_stop(sc);
866 	if (m0 != NULL)
867 		m_freem(m0);
868 
869 	RL_UNLOCK(sc);
870 
871 	return (error);
872 }
873 
874 #endif
875 
876 /*
877  * Probe for a RealTek 8139C+/8169/8110 chip. Check the PCI vendor and device
878  * IDs against our list and return a device name if we find a match.
879  */
880 static int
881 re_probe(device_t dev)
882 {
883 	struct rl_type		*t;
884 	uint16_t		devid, vendor;
885 	uint16_t		revid, sdevid;
886 	int			i;
887 
888 	vendor = pci_get_vendor(dev);
889 	devid = pci_get_device(dev);
890 	revid = pci_get_revid(dev);
891 	sdevid = pci_get_subdevice(dev);
892 
893 	if (vendor == LINKSYS_VENDORID && devid == LINKSYS_DEVICEID_EG1032) {
894 		if (sdevid != LINKSYS_SUBDEVICE_EG1032_REV3) {
895 			/*
896 			 * Only attach to rev. 3 of the Linksys EG1032 adapter.
897 			 * Rev. 2 is supported by sk(4).
898 			 */
899 			return (ENXIO);
900 		}
901 	}
902 
903 	if (vendor == RT_VENDORID && devid == RT_DEVICEID_8139) {
904 		if (revid != 0x20) {
905 			/* 8139, let rl(4) take care of this device. */
906 			return (ENXIO);
907 		}
908 	}
909 
910 	t = re_devs;
911 	for (i = 0; i < sizeof(re_devs) / sizeof(re_devs[0]); i++, t++) {
912 		if (vendor == t->rl_vid && devid == t->rl_did) {
913 			device_set_desc(dev, t->rl_name);
914 			return (BUS_PROBE_DEFAULT);
915 		}
916 	}
917 
918 	return (ENXIO);
919 }
920 
921 /*
922  * Map a single buffer address.
923  */
924 
925 static void
926 re_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
927 {
928 	bus_addr_t		*addr;
929 
930 	if (error)
931 		return;
932 
933 	KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg));
934 	addr = arg;
935 	*addr = segs->ds_addr;
936 }
937 
938 static int
939 re_allocmem(device_t dev, struct rl_softc *sc)
940 {
941 	bus_size_t		rx_list_size, tx_list_size;
942 	int			error;
943 	int			i;
944 
945 	rx_list_size = sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc);
946 	tx_list_size = sc->rl_ldata.rl_tx_desc_cnt * sizeof(struct rl_desc);
947 
948 	/*
949 	 * Allocate the parent bus DMA tag appropriate for PCI.
950 	 * In order to use DAC, RL_CPLUSCMD_PCI_DAC bit of RL_CPLUS_CMD
951 	 * register should be set. However some RealTek chips are known
952 	 * to be buggy on DAC handling, therefore disable DAC by limiting
953 	 * DMA address space to 32bit. PCIe variants of RealTek chips
954 	 * may not have the limitation but I took safer path.
955 	 */
956 	error = bus_dma_tag_create(bus_get_dma_tag(dev), 1, 0,
957 	    BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
958 	    BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0,
959 	    NULL, NULL, &sc->rl_parent_tag);
960 	if (error) {
961 		device_printf(dev, "could not allocate parent DMA tag\n");
962 		return (error);
963 	}
964 
965 	/*
966 	 * Allocate map for TX mbufs.
967 	 */
968 	error = bus_dma_tag_create(sc->rl_parent_tag, 1, 0,
969 	    BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
970 	    NULL, MCLBYTES * RL_NTXSEGS, RL_NTXSEGS, 4096, 0,
971 	    NULL, NULL, &sc->rl_ldata.rl_tx_mtag);
972 	if (error) {
973 		device_printf(dev, "could not allocate TX DMA tag\n");
974 		return (error);
975 	}
976 
977 	/*
978 	 * Allocate map for RX mbufs.
979 	 */
980 
981 	error = bus_dma_tag_create(sc->rl_parent_tag, sizeof(uint64_t), 0,
982 	    BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
983 	    MCLBYTES, 1, MCLBYTES, 0, NULL, NULL, &sc->rl_ldata.rl_rx_mtag);
984 	if (error) {
985 		device_printf(dev, "could not allocate RX DMA tag\n");
986 		return (error);
987 	}
988 
989 	/*
990 	 * Allocate map for TX descriptor list.
991 	 */
992 	error = bus_dma_tag_create(sc->rl_parent_tag, RL_RING_ALIGN,
993 	    0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL,
994 	    NULL, tx_list_size, 1, tx_list_size, 0,
995 	    NULL, NULL, &sc->rl_ldata.rl_tx_list_tag);
996 	if (error) {
997 		device_printf(dev, "could not allocate TX DMA ring tag\n");
998 		return (error);
999 	}
1000 
1001 	/* Allocate DMA'able memory for the TX ring */
1002 
1003 	error = bus_dmamem_alloc(sc->rl_ldata.rl_tx_list_tag,
1004 	    (void **)&sc->rl_ldata.rl_tx_list,
1005 	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
1006 	    &sc->rl_ldata.rl_tx_list_map);
1007 	if (error) {
1008 		device_printf(dev, "could not allocate TX DMA ring\n");
1009 		return (error);
1010 	}
1011 
1012 	/* Load the map for the TX ring. */
1013 
1014 	sc->rl_ldata.rl_tx_list_addr = 0;
1015 	error = bus_dmamap_load(sc->rl_ldata.rl_tx_list_tag,
1016 	     sc->rl_ldata.rl_tx_list_map, sc->rl_ldata.rl_tx_list,
1017 	     tx_list_size, re_dma_map_addr,
1018 	     &sc->rl_ldata.rl_tx_list_addr, BUS_DMA_NOWAIT);
1019 	if (error != 0 || sc->rl_ldata.rl_tx_list_addr == 0) {
1020 		device_printf(dev, "could not load TX DMA ring\n");
1021 		return (ENOMEM);
1022 	}
1023 
1024 	/* Create DMA maps for TX buffers */
1025 
1026 	for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) {
1027 		error = bus_dmamap_create(sc->rl_ldata.rl_tx_mtag, 0,
1028 		    &sc->rl_ldata.rl_tx_desc[i].tx_dmamap);
1029 		if (error) {
1030 			device_printf(dev, "could not create DMA map for TX\n");
1031 			return (error);
1032 		}
1033 	}
1034 
1035 	/*
1036 	 * Allocate map for RX descriptor list.
1037 	 */
1038 	error = bus_dma_tag_create(sc->rl_parent_tag, RL_RING_ALIGN,
1039 	    0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL,
1040 	    NULL, rx_list_size, 1, rx_list_size, 0,
1041 	    NULL, NULL, &sc->rl_ldata.rl_rx_list_tag);
1042 	if (error) {
1043 		device_printf(dev, "could not create RX DMA ring tag\n");
1044 		return (error);
1045 	}
1046 
1047 	/* Allocate DMA'able memory for the RX ring */
1048 
1049 	error = bus_dmamem_alloc(sc->rl_ldata.rl_rx_list_tag,
1050 	    (void **)&sc->rl_ldata.rl_rx_list,
1051 	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
1052 	    &sc->rl_ldata.rl_rx_list_map);
1053 	if (error) {
1054 		device_printf(dev, "could not allocate RX DMA ring\n");
1055 		return (error);
1056 	}
1057 
1058 	/* Load the map for the RX ring. */
1059 
1060 	sc->rl_ldata.rl_rx_list_addr = 0;
1061 	error = bus_dmamap_load(sc->rl_ldata.rl_rx_list_tag,
1062 	     sc->rl_ldata.rl_rx_list_map, sc->rl_ldata.rl_rx_list,
1063 	     rx_list_size, re_dma_map_addr,
1064 	     &sc->rl_ldata.rl_rx_list_addr, BUS_DMA_NOWAIT);
1065 	if (error != 0 || sc->rl_ldata.rl_rx_list_addr == 0) {
1066 		device_printf(dev, "could not load RX DMA ring\n");
1067 		return (ENOMEM);
1068 	}
1069 
1070 	/* Create DMA maps for RX buffers */
1071 
1072 	error = bus_dmamap_create(sc->rl_ldata.rl_rx_mtag, 0,
1073 	    &sc->rl_ldata.rl_rx_sparemap);
1074 	if (error) {
1075 		device_printf(dev, "could not create spare DMA map for RX\n");
1076 		return (error);
1077 	}
1078 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
1079 		error = bus_dmamap_create(sc->rl_ldata.rl_rx_mtag, 0,
1080 		    &sc->rl_ldata.rl_rx_desc[i].rx_dmamap);
1081 		if (error) {
1082 			device_printf(dev, "could not create DMA map for RX\n");
1083 			return (error);
1084 		}
1085 	}
1086 
1087 	return (0);
1088 }
1089 
1090 /*
1091  * Attach the interface. Allocate softc structures, do ifmedia
1092  * setup and ethernet/BPF attach.
1093  */
1094 static int
1095 re_attach(device_t dev)
1096 {
1097 	u_char			eaddr[ETHER_ADDR_LEN];
1098 	u_int16_t		as[ETHER_ADDR_LEN / 2];
1099 	struct rl_softc		*sc;
1100 	struct ifnet		*ifp;
1101 	struct rl_hwrev		*hw_rev;
1102 	int			hwrev;
1103 	u_int16_t		devid, re_did = 0;
1104 	int			error = 0, rid, i;
1105 	int			msic, reg;
1106 	uint8_t			cfg;
1107 
1108 	sc = device_get_softc(dev);
1109 	sc->rl_dev = dev;
1110 
1111 	mtx_init(&sc->rl_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
1112 	    MTX_DEF);
1113 	callout_init_mtx(&sc->rl_stat_callout, &sc->rl_mtx, 0);
1114 
1115 	/*
1116 	 * Map control/status registers.
1117 	 */
1118 	pci_enable_busmaster(dev);
1119 
1120 	devid = pci_get_device(dev);
1121 	/* Prefer memory space register mapping over IO space. */
1122 	sc->rl_res_id = PCIR_BAR(1);
1123 	sc->rl_res_type = SYS_RES_MEMORY;
1124 	/* RTL8168/8101E seems to use different BARs. */
1125 	if (devid == RT_DEVICEID_8168 || devid == RT_DEVICEID_8101E)
1126 		sc->rl_res_id = PCIR_BAR(2);
1127 	sc->rl_res = bus_alloc_resource_any(dev, sc->rl_res_type,
1128 	    &sc->rl_res_id, RF_ACTIVE);
1129 
1130 	if (sc->rl_res == NULL) {
1131 		sc->rl_res_id = PCIR_BAR(0);
1132 		sc->rl_res_type = SYS_RES_IOPORT;
1133 		sc->rl_res = bus_alloc_resource_any(dev, sc->rl_res_type,
1134 		    &sc->rl_res_id, RF_ACTIVE);
1135 		if (sc->rl_res == NULL) {
1136 			device_printf(dev, "couldn't map ports/memory\n");
1137 			error = ENXIO;
1138 			goto fail;
1139 		}
1140 	}
1141 
1142 	sc->rl_btag = rman_get_bustag(sc->rl_res);
1143 	sc->rl_bhandle = rman_get_bushandle(sc->rl_res);
1144 
1145 	msic = 0;
1146 	if (pci_find_extcap(dev, PCIY_EXPRESS, &reg) == 0) {
1147 		sc->rl_flags |= RL_FLAG_PCIE;
1148 		msic = pci_msi_count(dev);
1149 		if (bootverbose)
1150 			device_printf(dev, "MSI count : %d\n", msic);
1151 	}
1152 	if (msic > 0 && msi_disable == 0) {
1153 		msic = 1;
1154 		if (pci_alloc_msi(dev, &msic) == 0) {
1155 			if (msic == RL_MSI_MESSAGES) {
1156 				device_printf(dev, "Using %d MSI messages\n",
1157 				    msic);
1158 				sc->rl_flags |= RL_FLAG_MSI;
1159 				/* Explicitly set MSI enable bit. */
1160 				CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
1161 				cfg = CSR_READ_1(sc, RL_CFG2);
1162 				cfg |= RL_CFG2_MSI;
1163 				CSR_WRITE_1(sc, RL_CFG2, cfg);
1164 				CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
1165 			} else
1166 				pci_release_msi(dev);
1167 		}
1168 	}
1169 
1170 	/* Allocate interrupt */
1171 	if ((sc->rl_flags & RL_FLAG_MSI) == 0) {
1172 		rid = 0;
1173 		sc->rl_irq[0] = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
1174 		    RF_SHAREABLE | RF_ACTIVE);
1175 		if (sc->rl_irq[0] == NULL) {
1176 			device_printf(dev, "couldn't allocate IRQ resources\n");
1177 			error = ENXIO;
1178 			goto fail;
1179 		}
1180 	} else {
1181 		for (i = 0, rid = 1; i < RL_MSI_MESSAGES; i++, rid++) {
1182 			sc->rl_irq[i] = bus_alloc_resource_any(dev,
1183 			    SYS_RES_IRQ, &rid, RF_ACTIVE);
1184 			if (sc->rl_irq[i] == NULL) {
1185 				device_printf(dev,
1186 				    "couldn't llocate IRQ resources for "
1187 				    "message %d\n", rid);
1188 				error = ENXIO;
1189 				goto fail;
1190 			}
1191 		}
1192 	}
1193 
1194 	if ((sc->rl_flags & RL_FLAG_MSI) == 0) {
1195 		CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
1196 		cfg = CSR_READ_1(sc, RL_CFG2);
1197 		if ((cfg & RL_CFG2_MSI) != 0) {
1198 			device_printf(dev, "turning off MSI enable bit.\n");
1199 			cfg &= ~RL_CFG2_MSI;
1200 			CSR_WRITE_1(sc, RL_CFG2, cfg);
1201 		}
1202 		CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
1203 	}
1204 
1205 	/* Reset the adapter. */
1206 	RL_LOCK(sc);
1207 	re_reset(sc);
1208 	RL_UNLOCK(sc);
1209 
1210 	hw_rev = re_hwrevs;
1211 	hwrev = CSR_READ_4(sc, RL_TXCFG);
1212 	switch (hwrev & 0x70000000) {
1213 	case 0x00000000:
1214 	case 0x10000000:
1215 		device_printf(dev, "Chip rev. 0x%08x\n", hwrev & 0xfc800000);
1216 		hwrev &= (RL_TXCFG_HWREV | 0x80000000);
1217 		break;
1218 	default:
1219 		device_printf(dev, "Chip rev. 0x%08x\n", hwrev & 0x7c800000);
1220 		hwrev &= RL_TXCFG_HWREV;
1221 		break;
1222 	}
1223 	device_printf(dev, "MAC rev. 0x%08x\n", hwrev & 0x00700000);
1224 	while (hw_rev->rl_desc != NULL) {
1225 		if (hw_rev->rl_rev == hwrev) {
1226 			sc->rl_type = hw_rev->rl_type;
1227 			sc->rl_hwrev = hw_rev->rl_rev;
1228 			break;
1229 		}
1230 		hw_rev++;
1231 	}
1232 	if (hw_rev->rl_desc == NULL) {
1233 		device_printf(dev, "Unknown H/W revision: 0x%08x\n", hwrev);
1234 		error = ENXIO;
1235 		goto fail;
1236 	}
1237 
1238 	switch (hw_rev->rl_rev) {
1239 	case RL_HWREV_8139CPLUS:
1240 		sc->rl_flags |= RL_FLAG_NOJUMBO | RL_FLAG_FASTETHER;
1241 		break;
1242 	case RL_HWREV_8100E:
1243 	case RL_HWREV_8101E:
1244 		sc->rl_flags |= RL_FLAG_NOJUMBO | RL_FLAG_PHYWAKE |
1245 		    RL_FLAG_FASTETHER;
1246 		break;
1247 	case RL_HWREV_8102E:
1248 	case RL_HWREV_8102EL:
1249 		sc->rl_flags |= RL_FLAG_NOJUMBO | RL_FLAG_PHYWAKE |
1250 		    RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT |
1251 		    RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP;
1252 		break;
1253 	case RL_HWREV_8168_SPIN1:
1254 	case RL_HWREV_8168_SPIN2:
1255 		sc->rl_flags |= RL_FLAG_WOLRXENB;
1256 		/* FALLTHROUGH */
1257 	case RL_HWREV_8168_SPIN3:
1258 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_MACSTAT;
1259 		break;
1260 	case RL_HWREV_8168C_SPIN2:
1261 		sc->rl_flags |= RL_FLAG_MACSLEEP;
1262 		/* FALLTHROUGH */
1263 	case RL_HWREV_8168C:
1264 		if ((hwrev & 0x00700000) == 0x00200000)
1265 			sc->rl_flags |= RL_FLAG_MACSLEEP;
1266 		/* FALLTHROUGH */
1267 	case RL_HWREV_8168CP:
1268 	case RL_HWREV_8168D:
1269 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR |
1270 		    RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP;
1271 		/*
1272 		 * These controllers support jumbo frame but it seems
1273 		 * that enabling it requires touching additional magic
1274 		 * registers. Depending on MAC revisions some
1275 		 * controllers need to disable checksum offload. So
1276 		 * disable jumbo frame until I have better idea what
1277 		 * it really requires to make it support.
1278 		 * RTL8168C/CP : supports up to 6KB jumbo frame.
1279 		 * RTL8111C/CP : supports up to 9KB jumbo frame.
1280 		 */
1281 		sc->rl_flags |= RL_FLAG_NOJUMBO;
1282 		break;
1283 	case RL_HWREV_8169_8110SB:
1284 	case RL_HWREV_8169_8110SBL:
1285 	case RL_HWREV_8169_8110SC:
1286 	case RL_HWREV_8169_8110SCE:
1287 		sc->rl_flags |= RL_FLAG_PHYWAKE;
1288 		/* FALLTHROUGH */
1289 	case RL_HWREV_8169:
1290 	case RL_HWREV_8169S:
1291 	case RL_HWREV_8110S:
1292 		sc->rl_flags |= RL_FLAG_MACRESET;
1293 		break;
1294 	default:
1295 		break;
1296 	}
1297 
1298 	/* Enable PME. */
1299 	CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
1300 	cfg = CSR_READ_1(sc, RL_CFG1);
1301 	cfg |= RL_CFG1_PME;
1302 	CSR_WRITE_1(sc, RL_CFG1, cfg);
1303 	cfg = CSR_READ_1(sc, RL_CFG5);
1304 	cfg &= RL_CFG5_PME_STS;
1305 	CSR_WRITE_1(sc, RL_CFG5, cfg);
1306 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
1307 
1308 	if ((sc->rl_flags & RL_FLAG_PAR) != 0) {
1309 		/*
1310 		 * XXX Should have a better way to extract station
1311 		 * address from EEPROM.
1312 		 */
1313 		for (i = 0; i < ETHER_ADDR_LEN; i++)
1314 			eaddr[i] = CSR_READ_1(sc, RL_IDR0 + i);
1315 	} else {
1316 		sc->rl_eewidth = RL_9356_ADDR_LEN;
1317 		re_read_eeprom(sc, (caddr_t)&re_did, 0, 1);
1318 		if (re_did != 0x8129)
1319 			sc->rl_eewidth = RL_9346_ADDR_LEN;
1320 
1321 		/*
1322 		 * Get station address from the EEPROM.
1323 		 */
1324 		re_read_eeprom(sc, (caddr_t)as, RL_EE_EADDR, 3);
1325 		for (i = 0; i < ETHER_ADDR_LEN / 2; i++)
1326 			as[i] = le16toh(as[i]);
1327 		bcopy(as, eaddr, sizeof(eaddr));
1328 	}
1329 
1330 	if (sc->rl_type == RL_8169) {
1331 		/* Set RX length mask and number of descriptors. */
1332 		sc->rl_rxlenmask = RL_RDESC_STAT_GFRAGLEN;
1333 		sc->rl_txstart = RL_GTXSTART;
1334 		sc->rl_ldata.rl_tx_desc_cnt = RL_8169_TX_DESC_CNT;
1335 		sc->rl_ldata.rl_rx_desc_cnt = RL_8169_RX_DESC_CNT;
1336 	} else {
1337 		/* Set RX length mask and number of descriptors. */
1338 		sc->rl_rxlenmask = RL_RDESC_STAT_FRAGLEN;
1339 		sc->rl_txstart = RL_TXSTART;
1340 		sc->rl_ldata.rl_tx_desc_cnt = RL_8139_TX_DESC_CNT;
1341 		sc->rl_ldata.rl_rx_desc_cnt = RL_8139_RX_DESC_CNT;
1342 	}
1343 
1344 	error = re_allocmem(dev, sc);
1345 	if (error)
1346 		goto fail;
1347 
1348 	ifp = sc->rl_ifp = if_alloc(IFT_ETHER);
1349 	if (ifp == NULL) {
1350 		device_printf(dev, "can not if_alloc()\n");
1351 		error = ENOSPC;
1352 		goto fail;
1353 	}
1354 
1355 	/* Take controller out of deep sleep mode. */
1356 	if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) {
1357 		if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80)
1358 			CSR_WRITE_1(sc, RL_GPIO,
1359 			    CSR_READ_1(sc, RL_GPIO) | 0x01);
1360 		else
1361 			CSR_WRITE_1(sc, RL_GPIO,
1362 			    CSR_READ_1(sc, RL_GPIO) & ~0x01);
1363 	}
1364 
1365 	/* Take PHY out of power down mode. */
1366 	if ((sc->rl_flags & RL_FLAG_PHYWAKE) != 0) {
1367 		re_gmii_writereg(dev, 1, 0x1f, 0);
1368 		re_gmii_writereg(dev, 1, 0x0e, 0);
1369 	}
1370 
1371 	/* Do MII setup */
1372 	if (mii_phy_probe(dev, &sc->rl_miibus,
1373 	    re_ifmedia_upd, re_ifmedia_sts)) {
1374 		device_printf(dev, "MII without any phy!\n");
1375 		error = ENXIO;
1376 		goto fail;
1377 	}
1378 
1379 	ifp->if_softc = sc;
1380 	if_initname(ifp, device_get_name(dev), device_get_unit(dev));
1381 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
1382 	ifp->if_ioctl = re_ioctl;
1383 	ifp->if_start = re_start;
1384 	ifp->if_hwassist = RE_CSUM_FEATURES;
1385 	ifp->if_capabilities = IFCAP_HWCSUM;
1386 	ifp->if_capenable = ifp->if_capabilities;
1387 	ifp->if_init = re_init;
1388 	IFQ_SET_MAXLEN(&ifp->if_snd, RL_IFQ_MAXLEN);
1389 	ifp->if_snd.ifq_drv_maxlen = RL_IFQ_MAXLEN;
1390 	IFQ_SET_READY(&ifp->if_snd);
1391 
1392 	TASK_INIT(&sc->rl_txtask, 1, re_tx_task, ifp);
1393 	TASK_INIT(&sc->rl_inttask, 0, re_int_task, sc);
1394 
1395 	/*
1396 	 * XXX
1397 	 * Still have no idea how to make TSO work on 8168C, 8168CP,
1398 	 * 8111C and 8111CP.
1399 	 */
1400 	if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) {
1401 		ifp->if_hwassist |= CSUM_TSO;
1402 		ifp->if_capabilities |= IFCAP_TSO4;
1403 	}
1404 
1405 	/*
1406 	 * Call MI attach routine.
1407 	 */
1408 	ether_ifattach(ifp, eaddr);
1409 
1410 	/* VLAN capability setup */
1411 	ifp->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
1412 	if (ifp->if_capabilities & IFCAP_HWCSUM)
1413 		ifp->if_capabilities |= IFCAP_VLAN_HWCSUM;
1414 	/* Enable WOL if PM is supported. */
1415 	if (pci_find_extcap(sc->rl_dev, PCIY_PMG, &reg) == 0)
1416 		ifp->if_capabilities |= IFCAP_WOL;
1417 	ifp->if_capenable = ifp->if_capabilities;
1418 	/*
1419 	 * Don't enable TSO by default. Under certain
1420 	 * circumtances the controller generated corrupted
1421 	 * packets in TSO size.
1422 	 */
1423 	ifp->if_hwassist &= ~CSUM_TSO;
1424 	ifp->if_capenable &= ~IFCAP_TSO4;
1425 #ifdef DEVICE_POLLING
1426 	ifp->if_capabilities |= IFCAP_POLLING;
1427 #endif
1428 	/*
1429 	 * Tell the upper layer(s) we support long frames.
1430 	 * Must appear after the call to ether_ifattach() because
1431 	 * ether_ifattach() sets ifi_hdrlen to the default value.
1432 	 */
1433 	ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
1434 
1435 #ifdef RE_DIAG
1436 	/*
1437 	 * Perform hardware diagnostic on the original RTL8169.
1438 	 * Some 32-bit cards were incorrectly wired and would
1439 	 * malfunction if plugged into a 64-bit slot.
1440 	 */
1441 
1442 	if (hwrev == RL_HWREV_8169) {
1443 		error = re_diag(sc);
1444 		if (error) {
1445 			device_printf(dev,
1446 		    	"attach aborted due to hardware diag failure\n");
1447 			ether_ifdetach(ifp);
1448 			goto fail;
1449 		}
1450 	}
1451 #endif
1452 
1453 	/* Hook interrupt last to avoid having to lock softc */
1454 	if ((sc->rl_flags & RL_FLAG_MSI) == 0)
1455 		error = bus_setup_intr(dev, sc->rl_irq[0],
1456 		    INTR_TYPE_NET | INTR_MPSAFE, re_intr, NULL, sc,
1457 		    &sc->rl_intrhand[0]);
1458 	else {
1459 		for (i = 0; i < RL_MSI_MESSAGES; i++) {
1460 			error = bus_setup_intr(dev, sc->rl_irq[i],
1461 			    INTR_TYPE_NET | INTR_MPSAFE, re_intr, NULL, sc,
1462 		    	    &sc->rl_intrhand[i]);
1463 			if (error != 0)
1464 				break;
1465 		}
1466 	}
1467 	if (error) {
1468 		device_printf(dev, "couldn't set up irq\n");
1469 		ether_ifdetach(ifp);
1470 	}
1471 
1472 fail:
1473 
1474 	if (error)
1475 		re_detach(dev);
1476 
1477 	return (error);
1478 }
1479 
1480 /*
1481  * Shutdown hardware and free up resources. This can be called any
1482  * time after the mutex has been initialized. It is called in both
1483  * the error case in attach and the normal detach case so it needs
1484  * to be careful about only freeing resources that have actually been
1485  * allocated.
1486  */
1487 static int
1488 re_detach(device_t dev)
1489 {
1490 	struct rl_softc		*sc;
1491 	struct ifnet		*ifp;
1492 	int			i, rid;
1493 
1494 	sc = device_get_softc(dev);
1495 	ifp = sc->rl_ifp;
1496 	KASSERT(mtx_initialized(&sc->rl_mtx), ("re mutex not initialized"));
1497 
1498 	/* These should only be active if attach succeeded */
1499 	if (device_is_attached(dev)) {
1500 #ifdef DEVICE_POLLING
1501 		if (ifp->if_capenable & IFCAP_POLLING)
1502 			ether_poll_deregister(ifp);
1503 #endif
1504 		RL_LOCK(sc);
1505 #if 0
1506 		sc->suspended = 1;
1507 #endif
1508 		re_stop(sc);
1509 		RL_UNLOCK(sc);
1510 		callout_drain(&sc->rl_stat_callout);
1511 		taskqueue_drain(taskqueue_fast, &sc->rl_inttask);
1512 		taskqueue_drain(taskqueue_fast, &sc->rl_txtask);
1513 		/*
1514 		 * Force off the IFF_UP flag here, in case someone
1515 		 * still had a BPF descriptor attached to this
1516 		 * interface. If they do, ether_ifdetach() will cause
1517 		 * the BPF code to try and clear the promisc mode
1518 		 * flag, which will bubble down to re_ioctl(),
1519 		 * which will try to call re_init() again. This will
1520 		 * turn the NIC back on and restart the MII ticker,
1521 		 * which will panic the system when the kernel tries
1522 		 * to invoke the re_tick() function that isn't there
1523 		 * anymore.
1524 		 */
1525 		ifp->if_flags &= ~IFF_UP;
1526 		ether_ifdetach(ifp);
1527 	}
1528 	if (sc->rl_miibus)
1529 		device_delete_child(dev, sc->rl_miibus);
1530 	bus_generic_detach(dev);
1531 
1532 	/*
1533 	 * The rest is resource deallocation, so we should already be
1534 	 * stopped here.
1535 	 */
1536 
1537 	for (i = 0; i < RL_MSI_MESSAGES; i++) {
1538 		if (sc->rl_intrhand[i] != NULL) {
1539 			bus_teardown_intr(dev, sc->rl_irq[i],
1540 			    sc->rl_intrhand[i]);
1541 			sc->rl_intrhand[i] = NULL;
1542 		}
1543 	}
1544 	if (ifp != NULL)
1545 		if_free(ifp);
1546 	if ((sc->rl_flags & RL_FLAG_MSI) == 0) {
1547 		if (sc->rl_irq[0] != NULL) {
1548 			bus_release_resource(dev, SYS_RES_IRQ, 0,
1549 			    sc->rl_irq[0]);
1550 			sc->rl_irq[0] = NULL;
1551 		}
1552 	} else {
1553 		for (i = 0, rid = 1; i < RL_MSI_MESSAGES; i++, rid++) {
1554 			if (sc->rl_irq[i] != NULL) {
1555 				bus_release_resource(dev, SYS_RES_IRQ, rid,
1556 				    sc->rl_irq[i]);
1557 				sc->rl_irq[i] = NULL;
1558 			}
1559 		}
1560 		pci_release_msi(dev);
1561 	}
1562 	if (sc->rl_res)
1563 		bus_release_resource(dev, sc->rl_res_type, sc->rl_res_id,
1564 		    sc->rl_res);
1565 
1566 	/* Unload and free the RX DMA ring memory and map */
1567 
1568 	if (sc->rl_ldata.rl_rx_list_tag) {
1569 		bus_dmamap_unload(sc->rl_ldata.rl_rx_list_tag,
1570 		    sc->rl_ldata.rl_rx_list_map);
1571 		bus_dmamem_free(sc->rl_ldata.rl_rx_list_tag,
1572 		    sc->rl_ldata.rl_rx_list,
1573 		    sc->rl_ldata.rl_rx_list_map);
1574 		bus_dma_tag_destroy(sc->rl_ldata.rl_rx_list_tag);
1575 	}
1576 
1577 	/* Unload and free the TX DMA ring memory and map */
1578 
1579 	if (sc->rl_ldata.rl_tx_list_tag) {
1580 		bus_dmamap_unload(sc->rl_ldata.rl_tx_list_tag,
1581 		    sc->rl_ldata.rl_tx_list_map);
1582 		bus_dmamem_free(sc->rl_ldata.rl_tx_list_tag,
1583 		    sc->rl_ldata.rl_tx_list,
1584 		    sc->rl_ldata.rl_tx_list_map);
1585 		bus_dma_tag_destroy(sc->rl_ldata.rl_tx_list_tag);
1586 	}
1587 
1588 	/* Destroy all the RX and TX buffer maps */
1589 
1590 	if (sc->rl_ldata.rl_tx_mtag) {
1591 		for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++)
1592 			bus_dmamap_destroy(sc->rl_ldata.rl_tx_mtag,
1593 			    sc->rl_ldata.rl_tx_desc[i].tx_dmamap);
1594 		bus_dma_tag_destroy(sc->rl_ldata.rl_tx_mtag);
1595 	}
1596 	if (sc->rl_ldata.rl_rx_mtag) {
1597 		for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++)
1598 			bus_dmamap_destroy(sc->rl_ldata.rl_rx_mtag,
1599 			    sc->rl_ldata.rl_rx_desc[i].rx_dmamap);
1600 		if (sc->rl_ldata.rl_rx_sparemap)
1601 			bus_dmamap_destroy(sc->rl_ldata.rl_rx_mtag,
1602 			    sc->rl_ldata.rl_rx_sparemap);
1603 		bus_dma_tag_destroy(sc->rl_ldata.rl_rx_mtag);
1604 	}
1605 
1606 	/* Unload and free the stats buffer and map */
1607 
1608 	if (sc->rl_ldata.rl_stag) {
1609 		bus_dmamap_unload(sc->rl_ldata.rl_stag,
1610 		    sc->rl_ldata.rl_rx_list_map);
1611 		bus_dmamem_free(sc->rl_ldata.rl_stag,
1612 		    sc->rl_ldata.rl_stats,
1613 		    sc->rl_ldata.rl_smap);
1614 		bus_dma_tag_destroy(sc->rl_ldata.rl_stag);
1615 	}
1616 
1617 	if (sc->rl_parent_tag)
1618 		bus_dma_tag_destroy(sc->rl_parent_tag);
1619 
1620 	mtx_destroy(&sc->rl_mtx);
1621 
1622 	return (0);
1623 }
1624 
1625 static __inline void
1626 re_discard_rxbuf(struct rl_softc *sc, int idx)
1627 {
1628 	struct rl_desc		*desc;
1629 	struct rl_rxdesc	*rxd;
1630 	uint32_t		cmdstat;
1631 
1632 	rxd = &sc->rl_ldata.rl_rx_desc[idx];
1633 	desc = &sc->rl_ldata.rl_rx_list[idx];
1634 	desc->rl_vlanctl = 0;
1635 	cmdstat = rxd->rx_size;
1636 	if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1)
1637 		cmdstat |= RL_RDESC_CMD_EOR;
1638 	desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN);
1639 }
1640 
1641 static int
1642 re_newbuf(struct rl_softc *sc, int idx)
1643 {
1644 	struct mbuf		*m;
1645 	struct rl_rxdesc	*rxd;
1646 	bus_dma_segment_t	segs[1];
1647 	bus_dmamap_t		map;
1648 	struct rl_desc		*desc;
1649 	uint32_t		cmdstat;
1650 	int			error, nsegs;
1651 
1652 	m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
1653 	if (m == NULL)
1654 		return (ENOBUFS);
1655 
1656 	m->m_len = m->m_pkthdr.len = MCLBYTES;
1657 #ifdef RE_FIXUP_RX
1658 	/*
1659 	 * This is part of an evil trick to deal with non-x86 platforms.
1660 	 * The RealTek chip requires RX buffers to be aligned on 64-bit
1661 	 * boundaries, but that will hose non-x86 machines. To get around
1662 	 * this, we leave some empty space at the start of each buffer
1663 	 * and for non-x86 hosts, we copy the buffer back six bytes
1664 	 * to achieve word alignment. This is slightly more efficient
1665 	 * than allocating a new buffer, copying the contents, and
1666 	 * discarding the old buffer.
1667 	 */
1668 	m_adj(m, RE_ETHER_ALIGN);
1669 #endif
1670 	error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_rx_mtag,
1671 	    sc->rl_ldata.rl_rx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT);
1672 	if (error != 0) {
1673 		m_freem(m);
1674 		return (ENOBUFS);
1675 	}
1676 	KASSERT(nsegs == 1, ("%s: %d segment returned!", __func__, nsegs));
1677 
1678 	rxd = &sc->rl_ldata.rl_rx_desc[idx];
1679 	if (rxd->rx_m != NULL) {
1680 		bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap,
1681 		    BUS_DMASYNC_POSTREAD);
1682 		bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap);
1683 	}
1684 
1685 	rxd->rx_m = m;
1686 	map = rxd->rx_dmamap;
1687 	rxd->rx_dmamap = sc->rl_ldata.rl_rx_sparemap;
1688 	rxd->rx_size = segs[0].ds_len;
1689 	sc->rl_ldata.rl_rx_sparemap = map;
1690 	bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap,
1691 	    BUS_DMASYNC_PREREAD);
1692 
1693 	desc = &sc->rl_ldata.rl_rx_list[idx];
1694 	desc->rl_vlanctl = 0;
1695 	desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[0].ds_addr));
1696 	desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[0].ds_addr));
1697 	cmdstat = segs[0].ds_len;
1698 	if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1)
1699 		cmdstat |= RL_RDESC_CMD_EOR;
1700 	desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN);
1701 
1702 	return (0);
1703 }
1704 
1705 #ifdef RE_FIXUP_RX
1706 static __inline void
1707 re_fixup_rx(struct mbuf *m)
1708 {
1709 	int                     i;
1710 	uint16_t                *src, *dst;
1711 
1712 	src = mtod(m, uint16_t *);
1713 	dst = src - (RE_ETHER_ALIGN - ETHER_ALIGN) / sizeof *src;
1714 
1715 	for (i = 0; i < (m->m_len / sizeof(uint16_t) + 1); i++)
1716 		*dst++ = *src++;
1717 
1718 	m->m_data -= RE_ETHER_ALIGN - ETHER_ALIGN;
1719 }
1720 #endif
1721 
1722 static int
1723 re_tx_list_init(struct rl_softc *sc)
1724 {
1725 	struct rl_desc		*desc;
1726 	int			i;
1727 
1728 	RL_LOCK_ASSERT(sc);
1729 
1730 	bzero(sc->rl_ldata.rl_tx_list,
1731 	    sc->rl_ldata.rl_tx_desc_cnt * sizeof(struct rl_desc));
1732 	for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++)
1733 		sc->rl_ldata.rl_tx_desc[i].tx_m = NULL;
1734 	/* Set EOR. */
1735 	desc = &sc->rl_ldata.rl_tx_list[sc->rl_ldata.rl_tx_desc_cnt - 1];
1736 	desc->rl_cmdstat |= htole32(RL_TDESC_CMD_EOR);
1737 
1738 	bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag,
1739 	    sc->rl_ldata.rl_tx_list_map,
1740 	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
1741 
1742 	sc->rl_ldata.rl_tx_prodidx = 0;
1743 	sc->rl_ldata.rl_tx_considx = 0;
1744 	sc->rl_ldata.rl_tx_free = sc->rl_ldata.rl_tx_desc_cnt;
1745 
1746 	return (0);
1747 }
1748 
1749 static int
1750 re_rx_list_init(struct rl_softc *sc)
1751 {
1752 	int			error, i;
1753 
1754 	bzero(sc->rl_ldata.rl_rx_list,
1755 	    sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc));
1756 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
1757 		sc->rl_ldata.rl_rx_desc[i].rx_m = NULL;
1758 		if ((error = re_newbuf(sc, i)) != 0)
1759 			return (error);
1760 	}
1761 
1762 	/* Flush the RX descriptors */
1763 
1764 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
1765 	    sc->rl_ldata.rl_rx_list_map,
1766 	    BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD);
1767 
1768 	sc->rl_ldata.rl_rx_prodidx = 0;
1769 	sc->rl_head = sc->rl_tail = NULL;
1770 
1771 	return (0);
1772 }
1773 
1774 /*
1775  * RX handler for C+ and 8169. For the gigE chips, we support
1776  * the reception of jumbo frames that have been fragmented
1777  * across multiple 2K mbuf cluster buffers.
1778  */
1779 static int
1780 re_rxeof(struct rl_softc *sc)
1781 {
1782 	struct mbuf		*m;
1783 	struct ifnet		*ifp;
1784 	int			i, total_len;
1785 	struct rl_desc		*cur_rx;
1786 	u_int32_t		rxstat, rxvlan;
1787 	int			maxpkt = 16;
1788 
1789 	RL_LOCK_ASSERT(sc);
1790 
1791 	ifp = sc->rl_ifp;
1792 
1793 	/* Invalidate the descriptor memory */
1794 
1795 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
1796 	    sc->rl_ldata.rl_rx_list_map,
1797 	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
1798 
1799 	for (i = sc->rl_ldata.rl_rx_prodidx; maxpkt > 0;
1800 	    i = RL_RX_DESC_NXT(sc, i)) {
1801 		cur_rx = &sc->rl_ldata.rl_rx_list[i];
1802 		rxstat = le32toh(cur_rx->rl_cmdstat);
1803 		if ((rxstat & RL_RDESC_STAT_OWN) != 0)
1804 			break;
1805 		total_len = rxstat & sc->rl_rxlenmask;
1806 		rxvlan = le32toh(cur_rx->rl_vlanctl);
1807 		m = sc->rl_ldata.rl_rx_desc[i].rx_m;
1808 
1809 		if (!(rxstat & RL_RDESC_STAT_EOF)) {
1810 			if (re_newbuf(sc, i) != 0) {
1811 				/*
1812 				 * If this is part of a multi-fragment packet,
1813 				 * discard all the pieces.
1814 				 */
1815 				if (sc->rl_head != NULL) {
1816 					m_freem(sc->rl_head);
1817 					sc->rl_head = sc->rl_tail = NULL;
1818 				}
1819 				re_discard_rxbuf(sc, i);
1820 				continue;
1821 			}
1822 			m->m_len = RE_RX_DESC_BUFLEN;
1823 			if (sc->rl_head == NULL)
1824 				sc->rl_head = sc->rl_tail = m;
1825 			else {
1826 				m->m_flags &= ~M_PKTHDR;
1827 				sc->rl_tail->m_next = m;
1828 				sc->rl_tail = m;
1829 			}
1830 			continue;
1831 		}
1832 
1833 		/*
1834 		 * NOTE: for the 8139C+, the frame length field
1835 		 * is always 12 bits in size, but for the gigE chips,
1836 		 * it is 13 bits (since the max RX frame length is 16K).
1837 		 * Unfortunately, all 32 bits in the status word
1838 		 * were already used, so to make room for the extra
1839 		 * length bit, RealTek took out the 'frame alignment
1840 		 * error' bit and shifted the other status bits
1841 		 * over one slot. The OWN, EOR, FS and LS bits are
1842 		 * still in the same places. We have already extracted
1843 		 * the frame length and checked the OWN bit, so rather
1844 		 * than using an alternate bit mapping, we shift the
1845 		 * status bits one space to the right so we can evaluate
1846 		 * them using the 8169 status as though it was in the
1847 		 * same format as that of the 8139C+.
1848 		 */
1849 		if (sc->rl_type == RL_8169)
1850 			rxstat >>= 1;
1851 
1852 		/*
1853 		 * if total_len > 2^13-1, both _RXERRSUM and _GIANT will be
1854 		 * set, but if CRC is clear, it will still be a valid frame.
1855 		 */
1856 		if (rxstat & RL_RDESC_STAT_RXERRSUM && !(total_len > 8191 &&
1857 		    (rxstat & RL_RDESC_STAT_ERRS) == RL_RDESC_STAT_GIANT)) {
1858 			ifp->if_ierrors++;
1859 			/*
1860 			 * If this is part of a multi-fragment packet,
1861 			 * discard all the pieces.
1862 			 */
1863 			if (sc->rl_head != NULL) {
1864 				m_freem(sc->rl_head);
1865 				sc->rl_head = sc->rl_tail = NULL;
1866 			}
1867 			re_discard_rxbuf(sc, i);
1868 			continue;
1869 		}
1870 
1871 		/*
1872 		 * If allocating a replacement mbuf fails,
1873 		 * reload the current one.
1874 		 */
1875 
1876 		if (re_newbuf(sc, i) != 0) {
1877 			ifp->if_iqdrops++;
1878 			if (sc->rl_head != NULL) {
1879 				m_freem(sc->rl_head);
1880 				sc->rl_head = sc->rl_tail = NULL;
1881 			}
1882 			re_discard_rxbuf(sc, i);
1883 			continue;
1884 		}
1885 
1886 		if (sc->rl_head != NULL) {
1887 			m->m_len = total_len % RE_RX_DESC_BUFLEN;
1888 			if (m->m_len == 0)
1889 				m->m_len = RE_RX_DESC_BUFLEN;
1890 			/*
1891 			 * Special case: if there's 4 bytes or less
1892 			 * in this buffer, the mbuf can be discarded:
1893 			 * the last 4 bytes is the CRC, which we don't
1894 			 * care about anyway.
1895 			 */
1896 			if (m->m_len <= ETHER_CRC_LEN) {
1897 				sc->rl_tail->m_len -=
1898 				    (ETHER_CRC_LEN - m->m_len);
1899 				m_freem(m);
1900 			} else {
1901 				m->m_len -= ETHER_CRC_LEN;
1902 				m->m_flags &= ~M_PKTHDR;
1903 				sc->rl_tail->m_next = m;
1904 			}
1905 			m = sc->rl_head;
1906 			sc->rl_head = sc->rl_tail = NULL;
1907 			m->m_pkthdr.len = total_len - ETHER_CRC_LEN;
1908 		} else
1909 			m->m_pkthdr.len = m->m_len =
1910 			    (total_len - ETHER_CRC_LEN);
1911 
1912 #ifdef RE_FIXUP_RX
1913 		re_fixup_rx(m);
1914 #endif
1915 		ifp->if_ipackets++;
1916 		m->m_pkthdr.rcvif = ifp;
1917 
1918 		/* Do RX checksumming if enabled */
1919 
1920 		if (ifp->if_capenable & IFCAP_RXCSUM) {
1921 			if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) {
1922 				/* Check IP header checksum */
1923 				if (rxstat & RL_RDESC_STAT_PROTOID)
1924 					m->m_pkthdr.csum_flags |=
1925 					    CSUM_IP_CHECKED;
1926 				if (!(rxstat & RL_RDESC_STAT_IPSUMBAD))
1927 					m->m_pkthdr.csum_flags |=
1928 					    CSUM_IP_VALID;
1929 
1930 				/* Check TCP/UDP checksum */
1931 				if ((RL_TCPPKT(rxstat) &&
1932 				    !(rxstat & RL_RDESC_STAT_TCPSUMBAD)) ||
1933 				    (RL_UDPPKT(rxstat) &&
1934 				     !(rxstat & RL_RDESC_STAT_UDPSUMBAD))) {
1935 					m->m_pkthdr.csum_flags |=
1936 						CSUM_DATA_VALID|CSUM_PSEUDO_HDR;
1937 					m->m_pkthdr.csum_data = 0xffff;
1938 				}
1939 			} else {
1940 				/*
1941 				 * RTL8168C/RTL816CP/RTL8111C/RTL8111CP
1942 				 */
1943 				if ((rxstat & RL_RDESC_STAT_PROTOID) &&
1944 				    (rxvlan & RL_RDESC_IPV4))
1945 					m->m_pkthdr.csum_flags |=
1946 					    CSUM_IP_CHECKED;
1947 				if (!(rxstat & RL_RDESC_STAT_IPSUMBAD) &&
1948 				    (rxvlan & RL_RDESC_IPV4))
1949 					m->m_pkthdr.csum_flags |=
1950 					    CSUM_IP_VALID;
1951 				if (((rxstat & RL_RDESC_STAT_TCP) &&
1952 				    !(rxstat & RL_RDESC_STAT_TCPSUMBAD)) ||
1953 				    ((rxstat & RL_RDESC_STAT_UDP) &&
1954 				    !(rxstat & RL_RDESC_STAT_UDPSUMBAD))) {
1955 					m->m_pkthdr.csum_flags |=
1956 						CSUM_DATA_VALID|CSUM_PSEUDO_HDR;
1957 					m->m_pkthdr.csum_data = 0xffff;
1958 				}
1959 			}
1960 		}
1961 		maxpkt--;
1962 		if (rxvlan & RL_RDESC_VLANCTL_TAG) {
1963 			m->m_pkthdr.ether_vtag =
1964 			    bswap16((rxvlan & RL_RDESC_VLANCTL_DATA));
1965 			m->m_flags |= M_VLANTAG;
1966 		}
1967 		RL_UNLOCK(sc);
1968 		(*ifp->if_input)(ifp, m);
1969 		RL_LOCK(sc);
1970 	}
1971 
1972 	/* Flush the RX DMA ring */
1973 
1974 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
1975 	    sc->rl_ldata.rl_rx_list_map,
1976 	    BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD);
1977 
1978 	sc->rl_ldata.rl_rx_prodidx = i;
1979 
1980 	if (maxpkt)
1981 		return(EAGAIN);
1982 
1983 	return(0);
1984 }
1985 
1986 static void
1987 re_txeof(struct rl_softc *sc)
1988 {
1989 	struct ifnet		*ifp;
1990 	struct rl_txdesc	*txd;
1991 	u_int32_t		txstat;
1992 	int			cons;
1993 
1994 	cons = sc->rl_ldata.rl_tx_considx;
1995 	if (cons == sc->rl_ldata.rl_tx_prodidx)
1996 		return;
1997 
1998 	ifp = sc->rl_ifp;
1999 	/* Invalidate the TX descriptor list */
2000 	bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag,
2001 	    sc->rl_ldata.rl_tx_list_map,
2002 	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
2003 
2004 	for (; cons != sc->rl_ldata.rl_tx_prodidx;
2005 	    cons = RL_TX_DESC_NXT(sc, cons)) {
2006 		txstat = le32toh(sc->rl_ldata.rl_tx_list[cons].rl_cmdstat);
2007 		if (txstat & RL_TDESC_STAT_OWN)
2008 			break;
2009 		/*
2010 		 * We only stash mbufs in the last descriptor
2011 		 * in a fragment chain, which also happens to
2012 		 * be the only place where the TX status bits
2013 		 * are valid.
2014 		 */
2015 		if (txstat & RL_TDESC_CMD_EOF) {
2016 			txd = &sc->rl_ldata.rl_tx_desc[cons];
2017 			bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag,
2018 			    txd->tx_dmamap, BUS_DMASYNC_POSTWRITE);
2019 			bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag,
2020 			    txd->tx_dmamap);
2021 			KASSERT(txd->tx_m != NULL,
2022 			    ("%s: freeing NULL mbufs!", __func__));
2023 			m_freem(txd->tx_m);
2024 			txd->tx_m = NULL;
2025 			if (txstat & (RL_TDESC_STAT_EXCESSCOL|
2026 			    RL_TDESC_STAT_COLCNT))
2027 				ifp->if_collisions++;
2028 			if (txstat & RL_TDESC_STAT_TXERRSUM)
2029 				ifp->if_oerrors++;
2030 			else
2031 				ifp->if_opackets++;
2032 		}
2033 		sc->rl_ldata.rl_tx_free++;
2034 		ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2035 	}
2036 	sc->rl_ldata.rl_tx_considx = cons;
2037 
2038 	/* No changes made to the TX ring, so no flush needed */
2039 
2040 	if (sc->rl_ldata.rl_tx_free != sc->rl_ldata.rl_tx_desc_cnt) {
2041 #ifdef RE_TX_MODERATION
2042 		/*
2043 		 * If not all descriptors have been reaped yet, reload
2044 		 * the timer so that we will eventually get another
2045 		 * interrupt that will cause us to re-enter this routine.
2046 		 * This is done in case the transmitter has gone idle.
2047 		 */
2048 		CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2049 #endif
2050 	} else
2051 		sc->rl_watchdog_timer = 0;
2052 }
2053 
2054 static void
2055 re_tick(void *xsc)
2056 {
2057 	struct rl_softc		*sc;
2058 	struct mii_data		*mii;
2059 
2060 	sc = xsc;
2061 
2062 	RL_LOCK_ASSERT(sc);
2063 
2064 	mii = device_get_softc(sc->rl_miibus);
2065 	mii_tick(mii);
2066 	if ((sc->rl_flags & RL_FLAG_LINK) == 0)
2067 		re_miibus_statchg(sc->rl_dev);
2068 	/*
2069 	 * Reclaim transmitted frames here. Technically it is not
2070 	 * necessary to do here but it ensures periodic reclamation
2071 	 * regardless of Tx completion interrupt which seems to be
2072 	 * lost on PCIe based controllers under certain situations.
2073 	 */
2074 	re_txeof(sc);
2075 	re_watchdog(sc);
2076 	callout_reset(&sc->rl_stat_callout, hz, re_tick, sc);
2077 }
2078 
2079 #ifdef DEVICE_POLLING
2080 static void
2081 re_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
2082 {
2083 	struct rl_softc *sc = ifp->if_softc;
2084 
2085 	RL_LOCK(sc);
2086 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2087 		re_poll_locked(ifp, cmd, count);
2088 	RL_UNLOCK(sc);
2089 }
2090 
2091 static void
2092 re_poll_locked(struct ifnet *ifp, enum poll_cmd cmd, int count)
2093 {
2094 	struct rl_softc *sc = ifp->if_softc;
2095 
2096 	RL_LOCK_ASSERT(sc);
2097 
2098 	sc->rxcycles = count;
2099 	re_rxeof(sc);
2100 	re_txeof(sc);
2101 
2102 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2103 		taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_txtask);
2104 
2105 	if (cmd == POLL_AND_CHECK_STATUS) { /* also check status register */
2106 		u_int16_t       status;
2107 
2108 		status = CSR_READ_2(sc, RL_ISR);
2109 		if (status == 0xffff)
2110 			return;
2111 		if (status)
2112 			CSR_WRITE_2(sc, RL_ISR, status);
2113 		if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) &&
2114 		    (sc->rl_flags & RL_FLAG_PCIE))
2115 			CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2116 
2117 		/*
2118 		 * XXX check behaviour on receiver stalls.
2119 		 */
2120 
2121 		if (status & RL_ISR_SYSTEM_ERR)
2122 			re_init_locked(sc);
2123 	}
2124 }
2125 #endif /* DEVICE_POLLING */
2126 
2127 static int
2128 re_intr(void *arg)
2129 {
2130 	struct rl_softc		*sc;
2131 	uint16_t		status;
2132 
2133 	sc = arg;
2134 
2135 	status = CSR_READ_2(sc, RL_ISR);
2136 	if (status == 0xFFFF || (status & RL_INTRS_CPLUS) == 0)
2137                 return (FILTER_STRAY);
2138 	CSR_WRITE_2(sc, RL_IMR, 0);
2139 
2140 	taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_inttask);
2141 
2142 	return (FILTER_HANDLED);
2143 }
2144 
2145 static void
2146 re_int_task(void *arg, int npending)
2147 {
2148 	struct rl_softc		*sc;
2149 	struct ifnet		*ifp;
2150 	u_int16_t		status;
2151 	int			rval = 0;
2152 
2153 	sc = arg;
2154 	ifp = sc->rl_ifp;
2155 
2156 	RL_LOCK(sc);
2157 
2158 	status = CSR_READ_2(sc, RL_ISR);
2159         CSR_WRITE_2(sc, RL_ISR, status);
2160 
2161 	if (sc->suspended ||
2162 	    (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2163 		RL_UNLOCK(sc);
2164 		return;
2165 	}
2166 
2167 #ifdef DEVICE_POLLING
2168 	if  (ifp->if_capenable & IFCAP_POLLING) {
2169 		RL_UNLOCK(sc);
2170 		return;
2171 	}
2172 #endif
2173 
2174 	if (status & (RL_ISR_RX_OK|RL_ISR_RX_ERR|RL_ISR_FIFO_OFLOW))
2175 		rval = re_rxeof(sc);
2176 
2177 	/*
2178 	 * Some chips will ignore a second TX request issued
2179 	 * while an existing transmission is in progress. If
2180 	 * the transmitter goes idle but there are still
2181 	 * packets waiting to be sent, we need to restart the
2182 	 * channel here to flush them out. This only seems to
2183 	 * be required with the PCIe devices.
2184 	 */
2185 	if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) &&
2186 	    (sc->rl_flags & RL_FLAG_PCIE))
2187 		CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2188 	if (status & (
2189 #ifdef RE_TX_MODERATION
2190 	    RL_ISR_TIMEOUT_EXPIRED|
2191 #else
2192 	    RL_ISR_TX_OK|
2193 #endif
2194 	    RL_ISR_TX_ERR|RL_ISR_TX_DESC_UNAVAIL))
2195 		re_txeof(sc);
2196 
2197 	if (status & RL_ISR_SYSTEM_ERR)
2198 		re_init_locked(sc);
2199 
2200 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2201 		taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_txtask);
2202 
2203 	RL_UNLOCK(sc);
2204 
2205         if ((CSR_READ_2(sc, RL_ISR) & RL_INTRS_CPLUS) || rval) {
2206 		taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_inttask);
2207 		return;
2208 	}
2209 
2210 	CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS);
2211 }
2212 
2213 static int
2214 re_encap(struct rl_softc *sc, struct mbuf **m_head)
2215 {
2216 	struct rl_txdesc	*txd, *txd_last;
2217 	bus_dma_segment_t	segs[RL_NTXSEGS];
2218 	bus_dmamap_t		map;
2219 	struct mbuf		*m_new;
2220 	struct rl_desc		*desc;
2221 	int			nsegs, prod;
2222 	int			i, error, ei, si;
2223 	int			padlen;
2224 	uint32_t		cmdstat, csum_flags, vlanctl;
2225 
2226 	RL_LOCK_ASSERT(sc);
2227 	M_ASSERTPKTHDR((*m_head));
2228 
2229 	/*
2230 	 * With some of the RealTek chips, using the checksum offload
2231 	 * support in conjunction with the autopadding feature results
2232 	 * in the transmission of corrupt frames. For example, if we
2233 	 * need to send a really small IP fragment that's less than 60
2234 	 * bytes in size, and IP header checksumming is enabled, the
2235 	 * resulting ethernet frame that appears on the wire will
2236 	 * have garbled payload. To work around this, if TX IP checksum
2237 	 * offload is enabled, we always manually pad short frames out
2238 	 * to the minimum ethernet frame size.
2239 	 */
2240 	if ((sc->rl_flags & RL_FLAG_DESCV2) == 0 &&
2241 	    (*m_head)->m_pkthdr.len < RL_IP4CSUMTX_PADLEN &&
2242 	    ((*m_head)->m_pkthdr.csum_flags & CSUM_IP) != 0) {
2243 		padlen = RL_MIN_FRAMELEN - (*m_head)->m_pkthdr.len;
2244 		if (M_WRITABLE(*m_head) == 0) {
2245 			/* Get a writable copy. */
2246 			m_new = m_dup(*m_head, M_DONTWAIT);
2247 			m_freem(*m_head);
2248 			if (m_new == NULL) {
2249 				*m_head = NULL;
2250 				return (ENOBUFS);
2251 			}
2252 			*m_head = m_new;
2253 		}
2254 		if ((*m_head)->m_next != NULL ||
2255 		    M_TRAILINGSPACE(*m_head) < padlen) {
2256 			m_new = m_defrag(*m_head, M_DONTWAIT);
2257 			if (m_new == NULL) {
2258 				m_freem(*m_head);
2259 				*m_head = NULL;
2260 				return (ENOBUFS);
2261 			}
2262 		} else
2263 			m_new = *m_head;
2264 
2265 		/*
2266 		 * Manually pad short frames, and zero the pad space
2267 		 * to avoid leaking data.
2268 		 */
2269 		bzero(mtod(m_new, char *) + m_new->m_pkthdr.len, padlen);
2270 		m_new->m_pkthdr.len += padlen;
2271 		m_new->m_len = m_new->m_pkthdr.len;
2272 		*m_head = m_new;
2273 	}
2274 
2275 	prod = sc->rl_ldata.rl_tx_prodidx;
2276 	txd = &sc->rl_ldata.rl_tx_desc[prod];
2277 	error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap,
2278 	    *m_head, segs, &nsegs, BUS_DMA_NOWAIT);
2279 	if (error == EFBIG) {
2280 		m_new = m_collapse(*m_head, M_DONTWAIT, RL_NTXSEGS);
2281 		if (m_new == NULL) {
2282 			m_freem(*m_head);
2283 			*m_head = NULL;
2284 			return (ENOBUFS);
2285 		}
2286 		*m_head = m_new;
2287 		error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_tx_mtag,
2288 		    txd->tx_dmamap, *m_head, segs, &nsegs, BUS_DMA_NOWAIT);
2289 		if (error != 0) {
2290 			m_freem(*m_head);
2291 			*m_head = NULL;
2292 			return (error);
2293 		}
2294 	} else if (error != 0)
2295 		return (error);
2296 	if (nsegs == 0) {
2297 		m_freem(*m_head);
2298 		*m_head = NULL;
2299 		return (EIO);
2300 	}
2301 
2302 	/* Check for number of available descriptors. */
2303 	if (sc->rl_ldata.rl_tx_free - nsegs <= 1) {
2304 		bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap);
2305 		return (ENOBUFS);
2306 	}
2307 
2308 	bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap,
2309 	    BUS_DMASYNC_PREWRITE);
2310 
2311 	/*
2312 	 * Set up checksum offload. Note: checksum offload bits must
2313 	 * appear in all descriptors of a multi-descriptor transmit
2314 	 * attempt. This is according to testing done with an 8169
2315 	 * chip. This is a requirement.
2316 	 */
2317 	vlanctl = 0;
2318 	csum_flags = 0;
2319 	if (((*m_head)->m_pkthdr.csum_flags & CSUM_TSO) != 0)
2320 		csum_flags = RL_TDESC_CMD_LGSEND |
2321 		    ((uint32_t)(*m_head)->m_pkthdr.tso_segsz <<
2322 		    RL_TDESC_CMD_MSSVAL_SHIFT);
2323 	else {
2324 		/*
2325 		 * Unconditionally enable IP checksum if TCP or UDP
2326 		 * checksum is required. Otherwise, TCP/UDP checksum
2327 		 * does't make effects.
2328 		 */
2329 		if (((*m_head)->m_pkthdr.csum_flags & RE_CSUM_FEATURES) != 0) {
2330 			if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) {
2331 				csum_flags |= RL_TDESC_CMD_IPCSUM;
2332 				if (((*m_head)->m_pkthdr.csum_flags &
2333 				    CSUM_TCP) != 0)
2334 					csum_flags |= RL_TDESC_CMD_TCPCSUM;
2335 				if (((*m_head)->m_pkthdr.csum_flags &
2336 				    CSUM_UDP) != 0)
2337 					csum_flags |= RL_TDESC_CMD_UDPCSUM;
2338 			} else {
2339 				vlanctl |= RL_TDESC_CMD_IPCSUMV2;
2340 				if (((*m_head)->m_pkthdr.csum_flags &
2341 				    CSUM_TCP) != 0)
2342 					vlanctl |= RL_TDESC_CMD_TCPCSUMV2;
2343 				if (((*m_head)->m_pkthdr.csum_flags &
2344 				    CSUM_UDP) != 0)
2345 					vlanctl |= RL_TDESC_CMD_UDPCSUMV2;
2346 			}
2347 		}
2348 	}
2349 
2350 	/*
2351 	 * Set up hardware VLAN tagging. Note: vlan tag info must
2352 	 * appear in all descriptors of a multi-descriptor
2353 	 * transmission attempt.
2354 	 */
2355 	if ((*m_head)->m_flags & M_VLANTAG)
2356 		vlanctl |= bswap16((*m_head)->m_pkthdr.ether_vtag) |
2357 		    RL_TDESC_VLANCTL_TAG;
2358 
2359 	si = prod;
2360 	for (i = 0; i < nsegs; i++, prod = RL_TX_DESC_NXT(sc, prod)) {
2361 		desc = &sc->rl_ldata.rl_tx_list[prod];
2362 		desc->rl_vlanctl = htole32(vlanctl);
2363 		desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[i].ds_addr));
2364 		desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[i].ds_addr));
2365 		cmdstat = segs[i].ds_len;
2366 		if (i != 0)
2367 			cmdstat |= RL_TDESC_CMD_OWN;
2368 		if (prod == sc->rl_ldata.rl_tx_desc_cnt - 1)
2369 			cmdstat |= RL_TDESC_CMD_EOR;
2370 		desc->rl_cmdstat = htole32(cmdstat | csum_flags);
2371 		sc->rl_ldata.rl_tx_free--;
2372 	}
2373 	/* Update producer index. */
2374 	sc->rl_ldata.rl_tx_prodidx = prod;
2375 
2376 	/* Set EOF on the last descriptor. */
2377 	ei = RL_TX_DESC_PRV(sc, prod);
2378 	desc = &sc->rl_ldata.rl_tx_list[ei];
2379 	desc->rl_cmdstat |= htole32(RL_TDESC_CMD_EOF);
2380 
2381 	desc = &sc->rl_ldata.rl_tx_list[si];
2382 	/* Set SOF and transfer ownership of packet to the chip. */
2383 	desc->rl_cmdstat |= htole32(RL_TDESC_CMD_OWN | RL_TDESC_CMD_SOF);
2384 
2385 	/*
2386 	 * Insure that the map for this transmission
2387 	 * is placed at the array index of the last descriptor
2388 	 * in this chain.  (Swap last and first dmamaps.)
2389 	 */
2390 	txd_last = &sc->rl_ldata.rl_tx_desc[ei];
2391 	map = txd->tx_dmamap;
2392 	txd->tx_dmamap = txd_last->tx_dmamap;
2393 	txd_last->tx_dmamap = map;
2394 	txd_last->tx_m = *m_head;
2395 
2396 	return (0);
2397 }
2398 
2399 static void
2400 re_tx_task(void *arg, int npending)
2401 {
2402 	struct ifnet		*ifp;
2403 
2404 	ifp = arg;
2405 	re_start(ifp);
2406 }
2407 
2408 /*
2409  * Main transmit routine for C+ and gigE NICs.
2410  */
2411 static void
2412 re_start(struct ifnet *ifp)
2413 {
2414 	struct rl_softc		*sc;
2415 	struct mbuf		*m_head;
2416 	int			queued;
2417 
2418 	sc = ifp->if_softc;
2419 
2420 	RL_LOCK(sc);
2421 
2422 	if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
2423 	    IFF_DRV_RUNNING || (sc->rl_flags & RL_FLAG_LINK) == 0) {
2424 		RL_UNLOCK(sc);
2425 		return;
2426 	}
2427 
2428 	for (queued = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd) &&
2429 	    sc->rl_ldata.rl_tx_free > 1;) {
2430 		IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
2431 		if (m_head == NULL)
2432 			break;
2433 
2434 		if (re_encap(sc, &m_head) != 0) {
2435 			if (m_head == NULL)
2436 				break;
2437 			IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
2438 			ifp->if_drv_flags |= IFF_DRV_OACTIVE;
2439 			break;
2440 		}
2441 
2442 		/*
2443 		 * If there's a BPF listener, bounce a copy of this frame
2444 		 * to him.
2445 		 */
2446 		ETHER_BPF_MTAP(ifp, m_head);
2447 
2448 		queued++;
2449 	}
2450 
2451 	if (queued == 0) {
2452 #ifdef RE_TX_MODERATION
2453 		if (sc->rl_ldata.rl_tx_free != sc->rl_ldata.rl_tx_desc_cnt)
2454 			CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2455 #endif
2456 		RL_UNLOCK(sc);
2457 		return;
2458 	}
2459 
2460 	/* Flush the TX descriptors */
2461 
2462 	bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag,
2463 	    sc->rl_ldata.rl_tx_list_map,
2464 	    BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD);
2465 
2466 	CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2467 
2468 #ifdef RE_TX_MODERATION
2469 	/*
2470 	 * Use the countdown timer for interrupt moderation.
2471 	 * 'TX done' interrupts are disabled. Instead, we reset the
2472 	 * countdown timer, which will begin counting until it hits
2473 	 * the value in the TIMERINT register, and then trigger an
2474 	 * interrupt. Each time we write to the TIMERCNT register,
2475 	 * the timer count is reset to 0.
2476 	 */
2477 	CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2478 #endif
2479 
2480 	/*
2481 	 * Set a timeout in case the chip goes out to lunch.
2482 	 */
2483 	sc->rl_watchdog_timer = 5;
2484 
2485 	RL_UNLOCK(sc);
2486 }
2487 
2488 static void
2489 re_init(void *xsc)
2490 {
2491 	struct rl_softc		*sc = xsc;
2492 
2493 	RL_LOCK(sc);
2494 	re_init_locked(sc);
2495 	RL_UNLOCK(sc);
2496 }
2497 
2498 static void
2499 re_init_locked(struct rl_softc *sc)
2500 {
2501 	struct ifnet		*ifp = sc->rl_ifp;
2502 	struct mii_data		*mii;
2503 	uint32_t		reg;
2504 	uint16_t		cfg;
2505 	union {
2506 		uint32_t align_dummy;
2507 		u_char eaddr[ETHER_ADDR_LEN];
2508         } eaddr;
2509 
2510 	RL_LOCK_ASSERT(sc);
2511 
2512 	mii = device_get_softc(sc->rl_miibus);
2513 
2514 	/*
2515 	 * Cancel pending I/O and free all RX/TX buffers.
2516 	 */
2517 	re_stop(sc);
2518 
2519 	/* Put controller into known state. */
2520 	re_reset(sc);
2521 
2522 	/*
2523 	 * Enable C+ RX and TX mode, as well as VLAN stripping and
2524 	 * RX checksum offload. We must configure the C+ register
2525 	 * before all others.
2526 	 */
2527 	cfg = RL_CPLUSCMD_PCI_MRW;
2528 	if ((ifp->if_capenable & IFCAP_RXCSUM) != 0)
2529 		cfg |= RL_CPLUSCMD_RXCSUM_ENB;
2530 	if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
2531 		cfg |= RL_CPLUSCMD_VLANSTRIP;
2532 	if ((sc->rl_flags & RL_FLAG_MACSTAT) != 0) {
2533 		cfg |= RL_CPLUSCMD_MACSTAT_DIS;
2534 		/* XXX magic. */
2535 		cfg |= 0x0001;
2536 	} else
2537 		cfg |= RL_CPLUSCMD_RXENB | RL_CPLUSCMD_TXENB;
2538 	CSR_WRITE_2(sc, RL_CPLUS_CMD, cfg);
2539 	if (sc->rl_hwrev == RL_HWREV_8169_8110SC ||
2540 	    sc->rl_hwrev == RL_HWREV_8169_8110SCE) {
2541 		reg = 0x000fff00;
2542 		if ((CSR_READ_1(sc, RL_CFG2) & RL_CFG2_PCI66MHZ) != 0)
2543 			reg |= 0x000000ff;
2544 		if (sc->rl_hwrev == RL_HWREV_8169_8110SCE)
2545 			reg |= 0x00f00000;
2546 		CSR_WRITE_4(sc, 0x7c, reg);
2547 		/* Disable interrupt mitigation. */
2548 		CSR_WRITE_2(sc, 0xe2, 0);
2549 	}
2550 	/*
2551 	 * Disable TSO if interface MTU size is greater than MSS
2552 	 * allowed in controller.
2553 	 */
2554 	if (ifp->if_mtu > RL_TSO_MTU && (ifp->if_capenable & IFCAP_TSO4) != 0) {
2555 		ifp->if_capenable &= ~IFCAP_TSO4;
2556 		ifp->if_hwassist &= ~CSUM_TSO;
2557 	}
2558 
2559 	/*
2560 	 * Init our MAC address.  Even though the chipset
2561 	 * documentation doesn't mention it, we need to enter "Config
2562 	 * register write enable" mode to modify the ID registers.
2563 	 */
2564 	/* Copy MAC address on stack to align. */
2565 	bcopy(IF_LLADDR(ifp), eaddr.eaddr, ETHER_ADDR_LEN);
2566 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_WRITECFG);
2567 	CSR_WRITE_4(sc, RL_IDR0,
2568 	    htole32(*(u_int32_t *)(&eaddr.eaddr[0])));
2569 	CSR_WRITE_4(sc, RL_IDR4,
2570 	    htole32(*(u_int32_t *)(&eaddr.eaddr[4])));
2571 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
2572 
2573 	/*
2574 	 * For C+ mode, initialize the RX descriptors and mbufs.
2575 	 */
2576 	re_rx_list_init(sc);
2577 	re_tx_list_init(sc);
2578 
2579 	/*
2580 	 * Load the addresses of the RX and TX lists into the chip.
2581 	 */
2582 
2583 	CSR_WRITE_4(sc, RL_RXLIST_ADDR_HI,
2584 	    RL_ADDR_HI(sc->rl_ldata.rl_rx_list_addr));
2585 	CSR_WRITE_4(sc, RL_RXLIST_ADDR_LO,
2586 	    RL_ADDR_LO(sc->rl_ldata.rl_rx_list_addr));
2587 
2588 	CSR_WRITE_4(sc, RL_TXLIST_ADDR_HI,
2589 	    RL_ADDR_HI(sc->rl_ldata.rl_tx_list_addr));
2590 	CSR_WRITE_4(sc, RL_TXLIST_ADDR_LO,
2591 	    RL_ADDR_LO(sc->rl_ldata.rl_tx_list_addr));
2592 
2593 	/*
2594 	 * Enable transmit and receive.
2595 	 */
2596 	CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_TX_ENB|RL_CMD_RX_ENB);
2597 
2598 	/*
2599 	 * Set the initial TX configuration.
2600 	 */
2601 	if (sc->rl_testmode) {
2602 		if (sc->rl_type == RL_8169)
2603 			CSR_WRITE_4(sc, RL_TXCFG,
2604 			    RL_TXCFG_CONFIG|RL_LOOPTEST_ON);
2605 		else
2606 			CSR_WRITE_4(sc, RL_TXCFG,
2607 			    RL_TXCFG_CONFIG|RL_LOOPTEST_ON_CPLUS);
2608 	} else
2609 		CSR_WRITE_4(sc, RL_TXCFG, RL_TXCFG_CONFIG);
2610 
2611 	CSR_WRITE_1(sc, RL_EARLY_TX_THRESH, 16);
2612 
2613 	/*
2614 	 * Set the initial RX configuration.
2615 	 */
2616 	re_set_rxmode(sc);
2617 
2618 #ifdef DEVICE_POLLING
2619 	/*
2620 	 * Disable interrupts if we are polling.
2621 	 */
2622 	if (ifp->if_capenable & IFCAP_POLLING)
2623 		CSR_WRITE_2(sc, RL_IMR, 0);
2624 	else	/* otherwise ... */
2625 #endif
2626 
2627 	/*
2628 	 * Enable interrupts.
2629 	 */
2630 	if (sc->rl_testmode)
2631 		CSR_WRITE_2(sc, RL_IMR, 0);
2632 	else
2633 		CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS);
2634 	CSR_WRITE_2(sc, RL_ISR, RL_INTRS_CPLUS);
2635 
2636 	/* Set initial TX threshold */
2637 	sc->rl_txthresh = RL_TX_THRESH_INIT;
2638 
2639 	/* Start RX/TX process. */
2640 	CSR_WRITE_4(sc, RL_MISSEDPKT, 0);
2641 #ifdef notdef
2642 	/* Enable receiver and transmitter. */
2643 	CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_TX_ENB|RL_CMD_RX_ENB);
2644 #endif
2645 
2646 #ifdef RE_TX_MODERATION
2647 	/*
2648 	 * Initialize the timer interrupt register so that
2649 	 * a timer interrupt will be generated once the timer
2650 	 * reaches a certain number of ticks. The timer is
2651 	 * reloaded on each transmit. This gives us TX interrupt
2652 	 * moderation, which dramatically improves TX frame rate.
2653 	 */
2654 	if (sc->rl_type == RL_8169)
2655 		CSR_WRITE_4(sc, RL_TIMERINT_8169, 0x800);
2656 	else
2657 		CSR_WRITE_4(sc, RL_TIMERINT, 0x400);
2658 #endif
2659 
2660 	/*
2661 	 * For 8169 gigE NICs, set the max allowed RX packet
2662 	 * size so we can receive jumbo frames.
2663 	 */
2664 	if (sc->rl_type == RL_8169)
2665 		CSR_WRITE_2(sc, RL_MAXRXPKTLEN, 16383);
2666 
2667 	if (sc->rl_testmode)
2668 		return;
2669 
2670 	mii_mediachg(mii);
2671 
2672 	CSR_WRITE_1(sc, RL_CFG1, CSR_READ_1(sc, RL_CFG1) | RL_CFG1_DRVLOAD);
2673 
2674 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
2675 	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2676 
2677 	sc->rl_flags &= ~RL_FLAG_LINK;
2678 	sc->rl_watchdog_timer = 0;
2679 	callout_reset(&sc->rl_stat_callout, hz, re_tick, sc);
2680 }
2681 
2682 /*
2683  * Set media options.
2684  */
2685 static int
2686 re_ifmedia_upd(struct ifnet *ifp)
2687 {
2688 	struct rl_softc		*sc;
2689 	struct mii_data		*mii;
2690 	int			error;
2691 
2692 	sc = ifp->if_softc;
2693 	mii = device_get_softc(sc->rl_miibus);
2694 	RL_LOCK(sc);
2695 	error = mii_mediachg(mii);
2696 	RL_UNLOCK(sc);
2697 
2698 	return (error);
2699 }
2700 
2701 /*
2702  * Report current media status.
2703  */
2704 static void
2705 re_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
2706 {
2707 	struct rl_softc		*sc;
2708 	struct mii_data		*mii;
2709 
2710 	sc = ifp->if_softc;
2711 	mii = device_get_softc(sc->rl_miibus);
2712 
2713 	RL_LOCK(sc);
2714 	mii_pollstat(mii);
2715 	RL_UNLOCK(sc);
2716 	ifmr->ifm_active = mii->mii_media_active;
2717 	ifmr->ifm_status = mii->mii_media_status;
2718 }
2719 
2720 static int
2721 re_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
2722 {
2723 	struct rl_softc		*sc = ifp->if_softc;
2724 	struct ifreq		*ifr = (struct ifreq *) data;
2725 	struct mii_data		*mii;
2726 	int			error = 0;
2727 
2728 	switch (command) {
2729 	case SIOCSIFMTU:
2730 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > RL_JUMBO_MTU) {
2731 			error = EINVAL;
2732 			break;
2733 		}
2734 		if ((sc->rl_flags & RL_FLAG_NOJUMBO) != 0 &&
2735 		    ifr->ifr_mtu > RL_MAX_FRAMELEN) {
2736 			error = EINVAL;
2737 			break;
2738 		}
2739 		RL_LOCK(sc);
2740 		if (ifp->if_mtu != ifr->ifr_mtu)
2741 			ifp->if_mtu = ifr->ifr_mtu;
2742 		if (ifp->if_mtu > RL_TSO_MTU &&
2743 		    (ifp->if_capenable & IFCAP_TSO4) != 0) {
2744 			ifp->if_capenable &= ~IFCAP_TSO4;
2745 			ifp->if_hwassist &= ~CSUM_TSO;
2746 		}
2747 		RL_UNLOCK(sc);
2748 		break;
2749 	case SIOCSIFFLAGS:
2750 		RL_LOCK(sc);
2751 		if ((ifp->if_flags & IFF_UP) != 0) {
2752 			if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
2753 				if (((ifp->if_flags ^ sc->rl_if_flags)
2754 				    & (IFF_PROMISC | IFF_ALLMULTI)) != 0)
2755 					re_set_rxmode(sc);
2756 			} else
2757 				re_init_locked(sc);
2758 		} else {
2759 			if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
2760 				re_stop(sc);
2761 		}
2762 		sc->rl_if_flags = ifp->if_flags;
2763 		RL_UNLOCK(sc);
2764 		break;
2765 	case SIOCADDMULTI:
2766 	case SIOCDELMULTI:
2767 		RL_LOCK(sc);
2768 		re_set_rxmode(sc);
2769 		RL_UNLOCK(sc);
2770 		break;
2771 	case SIOCGIFMEDIA:
2772 	case SIOCSIFMEDIA:
2773 		mii = device_get_softc(sc->rl_miibus);
2774 		error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
2775 		break;
2776 	case SIOCSIFCAP:
2777 	    {
2778 		int mask, reinit;
2779 
2780 		mask = ifr->ifr_reqcap ^ ifp->if_capenable;
2781 		reinit = 0;
2782 #ifdef DEVICE_POLLING
2783 		if (mask & IFCAP_POLLING) {
2784 			if (ifr->ifr_reqcap & IFCAP_POLLING) {
2785 				error = ether_poll_register(re_poll, ifp);
2786 				if (error)
2787 					return(error);
2788 				RL_LOCK(sc);
2789 				/* Disable interrupts */
2790 				CSR_WRITE_2(sc, RL_IMR, 0x0000);
2791 				ifp->if_capenable |= IFCAP_POLLING;
2792 				RL_UNLOCK(sc);
2793 			} else {
2794 				error = ether_poll_deregister(ifp);
2795 				/* Enable interrupts. */
2796 				RL_LOCK(sc);
2797 				CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS);
2798 				ifp->if_capenable &= ~IFCAP_POLLING;
2799 				RL_UNLOCK(sc);
2800 			}
2801 		}
2802 #endif /* DEVICE_POLLING */
2803 		if (mask & IFCAP_HWCSUM) {
2804 			ifp->if_capenable ^= IFCAP_HWCSUM;
2805 			if (ifp->if_capenable & IFCAP_TXCSUM)
2806 				ifp->if_hwassist |= RE_CSUM_FEATURES;
2807 			else
2808 				ifp->if_hwassist &= ~RE_CSUM_FEATURES;
2809 			reinit = 1;
2810 		}
2811 		if (mask & IFCAP_VLAN_HWTAGGING) {
2812 			ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
2813 			reinit = 1;
2814 		}
2815 		if (mask & IFCAP_TSO4) {
2816 			ifp->if_capenable ^= IFCAP_TSO4;
2817 			if ((IFCAP_TSO4 & ifp->if_capenable) &&
2818 			    (IFCAP_TSO4 & ifp->if_capabilities))
2819 				ifp->if_hwassist |= CSUM_TSO;
2820 			else
2821 				ifp->if_hwassist &= ~CSUM_TSO;
2822 			if (ifp->if_mtu > RL_TSO_MTU &&
2823 			    (ifp->if_capenable & IFCAP_TSO4) != 0) {
2824 				ifp->if_capenable &= ~IFCAP_TSO4;
2825 				ifp->if_hwassist &= ~CSUM_TSO;
2826 			}
2827 		}
2828 		if ((mask & IFCAP_WOL) != 0 &&
2829 		    (ifp->if_capabilities & IFCAP_WOL) != 0) {
2830 			if ((mask & IFCAP_WOL_UCAST) != 0)
2831 				ifp->if_capenable ^= IFCAP_WOL_UCAST;
2832 			if ((mask & IFCAP_WOL_MCAST) != 0)
2833 				ifp->if_capenable ^= IFCAP_WOL_MCAST;
2834 			if ((mask & IFCAP_WOL_MAGIC) != 0)
2835 				ifp->if_capenable ^= IFCAP_WOL_MAGIC;
2836 		}
2837 		if (reinit && ifp->if_drv_flags & IFF_DRV_RUNNING)
2838 			re_init(sc);
2839 		VLAN_CAPABILITIES(ifp);
2840 	    }
2841 		break;
2842 	default:
2843 		error = ether_ioctl(ifp, command, data);
2844 		break;
2845 	}
2846 
2847 	return (error);
2848 }
2849 
2850 static void
2851 re_watchdog(struct rl_softc *sc)
2852 {
2853 	struct ifnet		*ifp;
2854 
2855 	RL_LOCK_ASSERT(sc);
2856 
2857 	if (sc->rl_watchdog_timer == 0 || --sc->rl_watchdog_timer != 0)
2858 		return;
2859 
2860 	ifp = sc->rl_ifp;
2861 	re_txeof(sc);
2862 	if (sc->rl_ldata.rl_tx_free == sc->rl_ldata.rl_tx_desc_cnt) {
2863 		if_printf(ifp, "watchdog timeout (missed Tx interrupts) "
2864 		    "-- recovering\n");
2865 		if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2866 			taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_txtask);
2867 		return;
2868 	}
2869 
2870 	if_printf(ifp, "watchdog timeout\n");
2871 	ifp->if_oerrors++;
2872 
2873 	re_rxeof(sc);
2874 	re_init_locked(sc);
2875 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2876 		taskqueue_enqueue_fast(taskqueue_fast, &sc->rl_txtask);
2877 }
2878 
2879 /*
2880  * Stop the adapter and free any mbufs allocated to the
2881  * RX and TX lists.
2882  */
2883 static void
2884 re_stop(struct rl_softc *sc)
2885 {
2886 	int			i;
2887 	struct ifnet		*ifp;
2888 	struct rl_txdesc	*txd;
2889 	struct rl_rxdesc	*rxd;
2890 
2891 	RL_LOCK_ASSERT(sc);
2892 
2893 	ifp = sc->rl_ifp;
2894 
2895 	sc->rl_watchdog_timer = 0;
2896 	callout_stop(&sc->rl_stat_callout);
2897 	ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
2898 
2899 	if ((sc->rl_flags & RL_FLAG_CMDSTOP) != 0)
2900 		CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_STOPREQ | RL_CMD_TX_ENB |
2901 		    RL_CMD_RX_ENB);
2902 	else
2903 		CSR_WRITE_1(sc, RL_COMMAND, 0x00);
2904 	DELAY(1000);
2905 	CSR_WRITE_2(sc, RL_IMR, 0x0000);
2906 	CSR_WRITE_2(sc, RL_ISR, 0xFFFF);
2907 
2908 	if (sc->rl_head != NULL) {
2909 		m_freem(sc->rl_head);
2910 		sc->rl_head = sc->rl_tail = NULL;
2911 	}
2912 
2913 	/* Free the TX list buffers. */
2914 
2915 	for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) {
2916 		txd = &sc->rl_ldata.rl_tx_desc[i];
2917 		if (txd->tx_m != NULL) {
2918 			bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag,
2919 			    txd->tx_dmamap, BUS_DMASYNC_POSTWRITE);
2920 			bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag,
2921 			    txd->tx_dmamap);
2922 			m_freem(txd->tx_m);
2923 			txd->tx_m = NULL;
2924 		}
2925 	}
2926 
2927 	/* Free the RX list buffers. */
2928 
2929 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
2930 		rxd = &sc->rl_ldata.rl_rx_desc[i];
2931 		if (rxd->rx_m != NULL) {
2932 			bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag,
2933 			    rxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
2934 			bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag,
2935 			    rxd->rx_dmamap);
2936 			m_freem(rxd->rx_m);
2937 			rxd->rx_m = NULL;
2938 		}
2939 	}
2940 }
2941 
2942 /*
2943  * Device suspend routine.  Stop the interface and save some PCI
2944  * settings in case the BIOS doesn't restore them properly on
2945  * resume.
2946  */
2947 static int
2948 re_suspend(device_t dev)
2949 {
2950 	struct rl_softc		*sc;
2951 
2952 	sc = device_get_softc(dev);
2953 
2954 	RL_LOCK(sc);
2955 	re_stop(sc);
2956 	re_setwol(sc);
2957 	sc->suspended = 1;
2958 	RL_UNLOCK(sc);
2959 
2960 	return (0);
2961 }
2962 
2963 /*
2964  * Device resume routine.  Restore some PCI settings in case the BIOS
2965  * doesn't, re-enable busmastering, and restart the interface if
2966  * appropriate.
2967  */
2968 static int
2969 re_resume(device_t dev)
2970 {
2971 	struct rl_softc		*sc;
2972 	struct ifnet		*ifp;
2973 
2974 	sc = device_get_softc(dev);
2975 
2976 	RL_LOCK(sc);
2977 
2978 	ifp = sc->rl_ifp;
2979 	/* Take controller out of sleep mode. */
2980 	if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) {
2981 		if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80)
2982 			CSR_WRITE_1(sc, RL_GPIO,
2983 			    CSR_READ_1(sc, RL_GPIO) | 0x01);
2984 	}
2985 
2986 	/* reinitialize interface if necessary */
2987 	if (ifp->if_flags & IFF_UP)
2988 		re_init_locked(sc);
2989 
2990 	/*
2991 	 * Clear WOL matching such that normal Rx filtering
2992 	 * wouldn't interfere with WOL patterns.
2993 	 */
2994 	re_clrwol(sc);
2995 	sc->suspended = 0;
2996 	RL_UNLOCK(sc);
2997 
2998 	return (0);
2999 }
3000 
3001 /*
3002  * Stop all chip I/O so that the kernel's probe routines don't
3003  * get confused by errant DMAs when rebooting.
3004  */
3005 static int
3006 re_shutdown(device_t dev)
3007 {
3008 	struct rl_softc		*sc;
3009 
3010 	sc = device_get_softc(dev);
3011 
3012 	RL_LOCK(sc);
3013 	re_stop(sc);
3014 	/*
3015 	 * Mark interface as down since otherwise we will panic if
3016 	 * interrupt comes in later on, which can happen in some
3017 	 * cases.
3018 	 */
3019 	sc->rl_ifp->if_flags &= ~IFF_UP;
3020 	re_setwol(sc);
3021 	RL_UNLOCK(sc);
3022 
3023 	return (0);
3024 }
3025 
3026 static void
3027 re_setwol(struct rl_softc *sc)
3028 {
3029 	struct ifnet		*ifp;
3030 	int			pmc;
3031 	uint16_t		pmstat;
3032 	uint8_t			v;
3033 
3034 	RL_LOCK_ASSERT(sc);
3035 
3036 	if (pci_find_extcap(sc->rl_dev, PCIY_PMG, &pmc) != 0)
3037 		return;
3038 
3039 	ifp = sc->rl_ifp;
3040 	/* Put controller into sleep mode. */
3041 	if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) {
3042 		if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80)
3043 			CSR_WRITE_1(sc, RL_GPIO,
3044 			    CSR_READ_1(sc, RL_GPIO) & ~0x01);
3045 	}
3046 	if ((ifp->if_capenable & IFCAP_WOL) != 0 &&
3047 	    (sc->rl_flags & RL_FLAG_WOLRXENB) != 0)
3048 		CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_RX_ENB);
3049 	/* Enable config register write. */
3050 	CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
3051 
3052 	/* Enable PME. */
3053 	v = CSR_READ_1(sc, RL_CFG1);
3054 	v &= ~RL_CFG1_PME;
3055 	if ((ifp->if_capenable & IFCAP_WOL) != 0)
3056 		v |= RL_CFG1_PME;
3057 	CSR_WRITE_1(sc, RL_CFG1, v);
3058 
3059 	v = CSR_READ_1(sc, RL_CFG3);
3060 	v &= ~(RL_CFG3_WOL_LINK | RL_CFG3_WOL_MAGIC);
3061 	if ((ifp->if_capenable & IFCAP_WOL_MAGIC) != 0)
3062 		v |= RL_CFG3_WOL_MAGIC;
3063 	CSR_WRITE_1(sc, RL_CFG3, v);
3064 
3065 	/* Config register write done. */
3066 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
3067 
3068 	v = CSR_READ_1(sc, RL_CFG5);
3069 	v &= ~(RL_CFG5_WOL_BCAST | RL_CFG5_WOL_MCAST | RL_CFG5_WOL_UCAST);
3070 	v &= ~RL_CFG5_WOL_LANWAKE;
3071 	if ((ifp->if_capenable & IFCAP_WOL_UCAST) != 0)
3072 		v |= RL_CFG5_WOL_UCAST;
3073 	if ((ifp->if_capenable & IFCAP_WOL_MCAST) != 0)
3074 		v |= RL_CFG5_WOL_MCAST | RL_CFG5_WOL_BCAST;
3075 	if ((ifp->if_capenable & IFCAP_WOL) != 0)
3076 		v |= RL_CFG5_WOL_LANWAKE;
3077 	CSR_WRITE_1(sc, RL_CFG5, v);
3078 
3079 	/*
3080 	 * It seems that hardware resets its link speed to 100Mbps in
3081 	 * power down mode so switching to 100Mbps in driver is not
3082 	 * needed.
3083 	 */
3084 
3085 	/* Request PME if WOL is requested. */
3086 	pmstat = pci_read_config(sc->rl_dev, pmc + PCIR_POWER_STATUS, 2);
3087 	pmstat &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3088 	if ((ifp->if_capenable & IFCAP_WOL) != 0)
3089 		pmstat |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3090 	pci_write_config(sc->rl_dev, pmc + PCIR_POWER_STATUS, pmstat, 2);
3091 }
3092 
3093 static void
3094 re_clrwol(struct rl_softc *sc)
3095 {
3096 	int			pmc;
3097 	uint8_t			v;
3098 
3099 	RL_LOCK_ASSERT(sc);
3100 
3101 	if (pci_find_extcap(sc->rl_dev, PCIY_PMG, &pmc) != 0)
3102 		return;
3103 
3104 	/* Enable config register write. */
3105 	CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
3106 
3107 	v = CSR_READ_1(sc, RL_CFG3);
3108 	v &= ~(RL_CFG3_WOL_LINK | RL_CFG3_WOL_MAGIC);
3109 	CSR_WRITE_1(sc, RL_CFG3, v);
3110 
3111 	/* Config register write done. */
3112 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
3113 
3114 	v = CSR_READ_1(sc, RL_CFG5);
3115 	v &= ~(RL_CFG5_WOL_BCAST | RL_CFG5_WOL_MCAST | RL_CFG5_WOL_UCAST);
3116 	v &= ~RL_CFG5_WOL_LANWAKE;
3117 	CSR_WRITE_1(sc, RL_CFG5, v);
3118 }
3119