xref: /freebsd/sys/dev/re/if_re.c (revision 4928135658a9d0eaee37003df6137ab363fcb0b4)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (c) 1997, 1998-2003
5  *	Bill Paul <wpaul@windriver.com>.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by Bill Paul.
18  * 4. Neither the name of the author nor the names of any co-contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
26  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32  * THE POSSIBILITY OF SUCH DAMAGE.
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 /*
39  * RealTek 8139C+/8169/8169S/8110S/8168/8111/8101E PCI NIC driver
40  *
41  * Written by Bill Paul <wpaul@windriver.com>
42  * Senior Networking Software Engineer
43  * Wind River Systems
44  */
45 
46 /*
47  * This driver is designed to support RealTek's next generation of
48  * 10/100 and 10/100/1000 PCI ethernet controllers. There are currently
49  * seven devices in this family: the RTL8139C+, the RTL8169, the RTL8169S,
50  * RTL8110S, the RTL8168, the RTL8111 and the RTL8101E.
51  *
52  * The 8139C+ is a 10/100 ethernet chip. It is backwards compatible
53  * with the older 8139 family, however it also supports a special
54  * C+ mode of operation that provides several new performance enhancing
55  * features. These include:
56  *
57  *	o Descriptor based DMA mechanism. Each descriptor represents
58  *	  a single packet fragment. Data buffers may be aligned on
59  *	  any byte boundary.
60  *
61  *	o 64-bit DMA
62  *
63  *	o TCP/IP checksum offload for both RX and TX
64  *
65  *	o High and normal priority transmit DMA rings
66  *
67  *	o VLAN tag insertion and extraction
68  *
69  *	o TCP large send (segmentation offload)
70  *
71  * Like the 8139, the 8139C+ also has a built-in 10/100 PHY. The C+
72  * programming API is fairly straightforward. The RX filtering, EEPROM
73  * access and PHY access is the same as it is on the older 8139 series
74  * chips.
75  *
76  * The 8169 is a 64-bit 10/100/1000 gigabit ethernet MAC. It has almost the
77  * same programming API and feature set as the 8139C+ with the following
78  * differences and additions:
79  *
80  *	o 1000Mbps mode
81  *
82  *	o Jumbo frames
83  *
84  *	o GMII and TBI ports/registers for interfacing with copper
85  *	  or fiber PHYs
86  *
87  *	o RX and TX DMA rings can have up to 1024 descriptors
88  *	  (the 8139C+ allows a maximum of 64)
89  *
90  *	o Slight differences in register layout from the 8139C+
91  *
92  * The TX start and timer interrupt registers are at different locations
93  * on the 8169 than they are on the 8139C+. Also, the status word in the
94  * RX descriptor has a slightly different bit layout. The 8169 does not
95  * have a built-in PHY. Most reference boards use a Marvell 88E1000 'Alaska'
96  * copper gigE PHY.
97  *
98  * The 8169S/8110S 10/100/1000 devices have built-in copper gigE PHYs
99  * (the 'S' stands for 'single-chip'). These devices have the same
100  * programming API as the older 8169, but also have some vendor-specific
101  * registers for the on-board PHY. The 8110S is a LAN-on-motherboard
102  * part designed to be pin-compatible with the RealTek 8100 10/100 chip.
103  *
104  * This driver takes advantage of the RX and TX checksum offload and
105  * VLAN tag insertion/extraction features. It also implements TX
106  * interrupt moderation using the timer interrupt registers, which
107  * significantly reduces TX interrupt load. There is also support
108  * for jumbo frames, however the 8169/8169S/8110S can not transmit
109  * jumbo frames larger than 7440, so the max MTU possible with this
110  * driver is 7422 bytes.
111  */
112 
113 #ifdef HAVE_KERNEL_OPTION_HEADERS
114 #include "opt_device_polling.h"
115 #endif
116 
117 #include <sys/param.h>
118 #include <sys/endian.h>
119 #include <sys/systm.h>
120 #include <sys/sockio.h>
121 #include <sys/mbuf.h>
122 #include <sys/malloc.h>
123 #include <sys/module.h>
124 #include <sys/kernel.h>
125 #include <sys/socket.h>
126 #include <sys/lock.h>
127 #include <sys/mutex.h>
128 #include <sys/sysctl.h>
129 #include <sys/taskqueue.h>
130 
131 #include <net/if.h>
132 #include <net/if_var.h>
133 #include <net/if_arp.h>
134 #include <net/ethernet.h>
135 #include <net/if_dl.h>
136 #include <net/if_media.h>
137 #include <net/if_types.h>
138 #include <net/if_vlan_var.h>
139 
140 #include <net/bpf.h>
141 
142 #include <netinet/netdump/netdump.h>
143 
144 #include <machine/bus.h>
145 #include <machine/resource.h>
146 #include <sys/bus.h>
147 #include <sys/rman.h>
148 
149 #include <dev/mii/mii.h>
150 #include <dev/mii/miivar.h>
151 
152 #include <dev/pci/pcireg.h>
153 #include <dev/pci/pcivar.h>
154 
155 #include <dev/rl/if_rlreg.h>
156 
157 MODULE_DEPEND(re, pci, 1, 1, 1);
158 MODULE_DEPEND(re, ether, 1, 1, 1);
159 MODULE_DEPEND(re, miibus, 1, 1, 1);
160 
161 /* "device miibus" required.  See GENERIC if you get errors here. */
162 #include "miibus_if.h"
163 
164 /* Tunables. */
165 static int intr_filter = 0;
166 TUNABLE_INT("hw.re.intr_filter", &intr_filter);
167 static int msi_disable = 0;
168 TUNABLE_INT("hw.re.msi_disable", &msi_disable);
169 static int msix_disable = 0;
170 TUNABLE_INT("hw.re.msix_disable", &msix_disable);
171 static int prefer_iomap = 0;
172 TUNABLE_INT("hw.re.prefer_iomap", &prefer_iomap);
173 
174 #define RE_CSUM_FEATURES    (CSUM_IP | CSUM_TCP | CSUM_UDP)
175 
176 /*
177  * Various supported device vendors/types and their names.
178  */
179 static const struct rl_type re_devs[] = {
180 	{ DLINK_VENDORID, DLINK_DEVICEID_528T, 0,
181 	    "D-Link DGE-528(T) Gigabit Ethernet Adapter" },
182 	{ DLINK_VENDORID, DLINK_DEVICEID_530T_REVC, 0,
183 	    "D-Link DGE-530(T) Gigabit Ethernet Adapter" },
184 	{ RT_VENDORID, RT_DEVICEID_8139, 0,
185 	    "RealTek 8139C+ 10/100BaseTX" },
186 	{ RT_VENDORID, RT_DEVICEID_8101E, 0,
187 	    "RealTek 810xE PCIe 10/100baseTX" },
188 	{ RT_VENDORID, RT_DEVICEID_8168, 0,
189 	    "RealTek 8168/8111 B/C/CP/D/DP/E/F/G PCIe Gigabit Ethernet" },
190 	{ NCUBE_VENDORID, RT_DEVICEID_8168, 0,
191 	    "TP-Link TG-3468 v2 (RTL8168) Gigabit Ethernet" },
192 	{ RT_VENDORID, RT_DEVICEID_8169, 0,
193 	    "RealTek 8169/8169S/8169SB(L)/8110S/8110SB(L) Gigabit Ethernet" },
194 	{ RT_VENDORID, RT_DEVICEID_8169SC, 0,
195 	    "RealTek 8169SC/8110SC Single-chip Gigabit Ethernet" },
196 	{ COREGA_VENDORID, COREGA_DEVICEID_CGLAPCIGT, 0,
197 	    "Corega CG-LAPCIGT (RTL8169S) Gigabit Ethernet" },
198 	{ LINKSYS_VENDORID, LINKSYS_DEVICEID_EG1032, 0,
199 	    "Linksys EG1032 (RTL8169S) Gigabit Ethernet" },
200 	{ USR_VENDORID, USR_DEVICEID_997902, 0,
201 	    "US Robotics 997902 (RTL8169S) Gigabit Ethernet" }
202 };
203 
204 static const struct rl_hwrev re_hwrevs[] = {
205 	{ RL_HWREV_8139, RL_8139, "", RL_MTU },
206 	{ RL_HWREV_8139A, RL_8139, "A", RL_MTU },
207 	{ RL_HWREV_8139AG, RL_8139, "A-G", RL_MTU },
208 	{ RL_HWREV_8139B, RL_8139, "B", RL_MTU },
209 	{ RL_HWREV_8130, RL_8139, "8130", RL_MTU },
210 	{ RL_HWREV_8139C, RL_8139, "C", RL_MTU },
211 	{ RL_HWREV_8139D, RL_8139, "8139D/8100B/8100C", RL_MTU },
212 	{ RL_HWREV_8139CPLUS, RL_8139CPLUS, "C+", RL_MTU },
213 	{ RL_HWREV_8168B_SPIN1, RL_8169, "8168", RL_JUMBO_MTU },
214 	{ RL_HWREV_8169, RL_8169, "8169", RL_JUMBO_MTU },
215 	{ RL_HWREV_8169S, RL_8169, "8169S", RL_JUMBO_MTU },
216 	{ RL_HWREV_8110S, RL_8169, "8110S", RL_JUMBO_MTU },
217 	{ RL_HWREV_8169_8110SB, RL_8169, "8169SB/8110SB", RL_JUMBO_MTU },
218 	{ RL_HWREV_8169_8110SC, RL_8169, "8169SC/8110SC", RL_JUMBO_MTU },
219 	{ RL_HWREV_8169_8110SBL, RL_8169, "8169SBL/8110SBL", RL_JUMBO_MTU },
220 	{ RL_HWREV_8169_8110SCE, RL_8169, "8169SC/8110SC", RL_JUMBO_MTU },
221 	{ RL_HWREV_8100, RL_8139, "8100", RL_MTU },
222 	{ RL_HWREV_8101, RL_8139, "8101", RL_MTU },
223 	{ RL_HWREV_8100E, RL_8169, "8100E", RL_MTU },
224 	{ RL_HWREV_8101E, RL_8169, "8101E", RL_MTU },
225 	{ RL_HWREV_8102E, RL_8169, "8102E", RL_MTU },
226 	{ RL_HWREV_8102EL, RL_8169, "8102EL", RL_MTU },
227 	{ RL_HWREV_8102EL_SPIN1, RL_8169, "8102EL", RL_MTU },
228 	{ RL_HWREV_8103E, RL_8169, "8103E", RL_MTU },
229 	{ RL_HWREV_8401E, RL_8169, "8401E", RL_MTU },
230 	{ RL_HWREV_8402, RL_8169, "8402", RL_MTU },
231 	{ RL_HWREV_8105E, RL_8169, "8105E", RL_MTU },
232 	{ RL_HWREV_8105E_SPIN1, RL_8169, "8105E", RL_MTU },
233 	{ RL_HWREV_8106E, RL_8169, "8106E", RL_MTU },
234 	{ RL_HWREV_8168B_SPIN2, RL_8169, "8168", RL_JUMBO_MTU },
235 	{ RL_HWREV_8168B_SPIN3, RL_8169, "8168", RL_JUMBO_MTU },
236 	{ RL_HWREV_8168C, RL_8169, "8168C/8111C", RL_JUMBO_MTU_6K },
237 	{ RL_HWREV_8168C_SPIN2, RL_8169, "8168C/8111C", RL_JUMBO_MTU_6K },
238 	{ RL_HWREV_8168CP, RL_8169, "8168CP/8111CP", RL_JUMBO_MTU_6K },
239 	{ RL_HWREV_8168D, RL_8169, "8168D/8111D", RL_JUMBO_MTU_9K },
240 	{ RL_HWREV_8168DP, RL_8169, "8168DP/8111DP", RL_JUMBO_MTU_9K },
241 	{ RL_HWREV_8168E, RL_8169, "8168E/8111E", RL_JUMBO_MTU_9K},
242 	{ RL_HWREV_8168E_VL, RL_8169, "8168E/8111E-VL", RL_JUMBO_MTU_6K},
243 	{ RL_HWREV_8168EP, RL_8169, "8168EP/8111EP", RL_JUMBO_MTU_9K},
244 	{ RL_HWREV_8168F, RL_8169, "8168F/8111F", RL_JUMBO_MTU_9K},
245 	{ RL_HWREV_8168G, RL_8169, "8168G/8111G", RL_JUMBO_MTU_9K},
246 	{ RL_HWREV_8168GU, RL_8169, "8168GU/8111GU", RL_JUMBO_MTU_9K},
247 	{ RL_HWREV_8168H, RL_8169, "8168H/8111H", RL_JUMBO_MTU_9K},
248 	{ RL_HWREV_8411, RL_8169, "8411", RL_JUMBO_MTU_9K},
249 	{ RL_HWREV_8411B, RL_8169, "8411B", RL_JUMBO_MTU_9K},
250 	{ 0, 0, NULL, 0 }
251 };
252 
253 static int re_probe		(device_t);
254 static int re_attach		(device_t);
255 static int re_detach		(device_t);
256 
257 static int re_encap		(struct rl_softc *, struct mbuf **);
258 
259 static void re_dma_map_addr	(void *, bus_dma_segment_t *, int, int);
260 static int re_allocmem		(device_t, struct rl_softc *);
261 static __inline void re_discard_rxbuf
262 				(struct rl_softc *, int);
263 static int re_newbuf		(struct rl_softc *, int);
264 static int re_jumbo_newbuf	(struct rl_softc *, int);
265 static int re_rx_list_init	(struct rl_softc *);
266 static int re_jrx_list_init	(struct rl_softc *);
267 static int re_tx_list_init	(struct rl_softc *);
268 #ifdef RE_FIXUP_RX
269 static __inline void re_fixup_rx
270 				(struct mbuf *);
271 #endif
272 static int re_rxeof		(struct rl_softc *, int *);
273 static void re_txeof		(struct rl_softc *);
274 #ifdef DEVICE_POLLING
275 static int re_poll		(struct ifnet *, enum poll_cmd, int);
276 static int re_poll_locked	(struct ifnet *, enum poll_cmd, int);
277 #endif
278 static int re_intr		(void *);
279 static void re_intr_msi		(void *);
280 static void re_tick		(void *);
281 static void re_int_task		(void *, int);
282 static void re_start		(struct ifnet *);
283 static void re_start_locked	(struct ifnet *);
284 static void re_start_tx		(struct rl_softc *);
285 static int re_ioctl		(struct ifnet *, u_long, caddr_t);
286 static void re_init		(void *);
287 static void re_init_locked	(struct rl_softc *);
288 static void re_stop		(struct rl_softc *);
289 static void re_watchdog		(struct rl_softc *);
290 static int re_suspend		(device_t);
291 static int re_resume		(device_t);
292 static int re_shutdown		(device_t);
293 static int re_ifmedia_upd	(struct ifnet *);
294 static void re_ifmedia_sts	(struct ifnet *, struct ifmediareq *);
295 
296 static void re_eeprom_putbyte	(struct rl_softc *, int);
297 static void re_eeprom_getword	(struct rl_softc *, int, u_int16_t *);
298 static void re_read_eeprom	(struct rl_softc *, caddr_t, int, int);
299 static int re_gmii_readreg	(device_t, int, int);
300 static int re_gmii_writereg	(device_t, int, int, int);
301 
302 static int re_miibus_readreg	(device_t, int, int);
303 static int re_miibus_writereg	(device_t, int, int, int);
304 static void re_miibus_statchg	(device_t);
305 
306 static void re_set_jumbo	(struct rl_softc *, int);
307 static void re_set_rxmode		(struct rl_softc *);
308 static void re_reset		(struct rl_softc *);
309 static void re_setwol		(struct rl_softc *);
310 static void re_clrwol		(struct rl_softc *);
311 static void re_set_linkspeed	(struct rl_softc *);
312 
313 NETDUMP_DEFINE(re);
314 
315 #ifdef DEV_NETMAP	/* see ixgbe.c for details */
316 #include <dev/netmap/if_re_netmap.h>
317 MODULE_DEPEND(re, netmap, 1, 1, 1);
318 #endif /* !DEV_NETMAP */
319 
320 #ifdef RE_DIAG
321 static int re_diag		(struct rl_softc *);
322 #endif
323 
324 static void re_add_sysctls	(struct rl_softc *);
325 static int re_sysctl_stats	(SYSCTL_HANDLER_ARGS);
326 static int sysctl_int_range	(SYSCTL_HANDLER_ARGS, int, int);
327 static int sysctl_hw_re_int_mod	(SYSCTL_HANDLER_ARGS);
328 
329 static device_method_t re_methods[] = {
330 	/* Device interface */
331 	DEVMETHOD(device_probe,		re_probe),
332 	DEVMETHOD(device_attach,	re_attach),
333 	DEVMETHOD(device_detach,	re_detach),
334 	DEVMETHOD(device_suspend,	re_suspend),
335 	DEVMETHOD(device_resume,	re_resume),
336 	DEVMETHOD(device_shutdown,	re_shutdown),
337 
338 	/* MII interface */
339 	DEVMETHOD(miibus_readreg,	re_miibus_readreg),
340 	DEVMETHOD(miibus_writereg,	re_miibus_writereg),
341 	DEVMETHOD(miibus_statchg,	re_miibus_statchg),
342 
343 	DEVMETHOD_END
344 };
345 
346 static driver_t re_driver = {
347 	"re",
348 	re_methods,
349 	sizeof(struct rl_softc)
350 };
351 
352 static devclass_t re_devclass;
353 
354 DRIVER_MODULE(re, pci, re_driver, re_devclass, 0, 0);
355 DRIVER_MODULE(miibus, re, miibus_driver, miibus_devclass, 0, 0);
356 
357 #define EE_SET(x)					\
358 	CSR_WRITE_1(sc, RL_EECMD,			\
359 		CSR_READ_1(sc, RL_EECMD) | x)
360 
361 #define EE_CLR(x)					\
362 	CSR_WRITE_1(sc, RL_EECMD,			\
363 		CSR_READ_1(sc, RL_EECMD) & ~x)
364 
365 /*
366  * Send a read command and address to the EEPROM, check for ACK.
367  */
368 static void
369 re_eeprom_putbyte(struct rl_softc *sc, int addr)
370 {
371 	int			d, i;
372 
373 	d = addr | (RL_9346_READ << sc->rl_eewidth);
374 
375 	/*
376 	 * Feed in each bit and strobe the clock.
377 	 */
378 
379 	for (i = 1 << (sc->rl_eewidth + 3); i; i >>= 1) {
380 		if (d & i) {
381 			EE_SET(RL_EE_DATAIN);
382 		} else {
383 			EE_CLR(RL_EE_DATAIN);
384 		}
385 		DELAY(100);
386 		EE_SET(RL_EE_CLK);
387 		DELAY(150);
388 		EE_CLR(RL_EE_CLK);
389 		DELAY(100);
390 	}
391 }
392 
393 /*
394  * Read a word of data stored in the EEPROM at address 'addr.'
395  */
396 static void
397 re_eeprom_getword(struct rl_softc *sc, int addr, u_int16_t *dest)
398 {
399 	int			i;
400 	u_int16_t		word = 0;
401 
402 	/*
403 	 * Send address of word we want to read.
404 	 */
405 	re_eeprom_putbyte(sc, addr);
406 
407 	/*
408 	 * Start reading bits from EEPROM.
409 	 */
410 	for (i = 0x8000; i; i >>= 1) {
411 		EE_SET(RL_EE_CLK);
412 		DELAY(100);
413 		if (CSR_READ_1(sc, RL_EECMD) & RL_EE_DATAOUT)
414 			word |= i;
415 		EE_CLR(RL_EE_CLK);
416 		DELAY(100);
417 	}
418 
419 	*dest = word;
420 }
421 
422 /*
423  * Read a sequence of words from the EEPROM.
424  */
425 static void
426 re_read_eeprom(struct rl_softc *sc, caddr_t dest, int off, int cnt)
427 {
428 	int			i;
429 	u_int16_t		word = 0, *ptr;
430 
431 	CSR_SETBIT_1(sc, RL_EECMD, RL_EEMODE_PROGRAM);
432 
433         DELAY(100);
434 
435 	for (i = 0; i < cnt; i++) {
436 		CSR_SETBIT_1(sc, RL_EECMD, RL_EE_SEL);
437 		re_eeprom_getword(sc, off + i, &word);
438 		CSR_CLRBIT_1(sc, RL_EECMD, RL_EE_SEL);
439 		ptr = (u_int16_t *)(dest + (i * 2));
440                 *ptr = word;
441 	}
442 
443 	CSR_CLRBIT_1(sc, RL_EECMD, RL_EEMODE_PROGRAM);
444 }
445 
446 static int
447 re_gmii_readreg(device_t dev, int phy, int reg)
448 {
449 	struct rl_softc		*sc;
450 	u_int32_t		rval;
451 	int			i;
452 
453 	sc = device_get_softc(dev);
454 
455 	/* Let the rgephy driver read the GMEDIASTAT register */
456 
457 	if (reg == RL_GMEDIASTAT) {
458 		rval = CSR_READ_1(sc, RL_GMEDIASTAT);
459 		return (rval);
460 	}
461 
462 	CSR_WRITE_4(sc, RL_PHYAR, reg << 16);
463 
464 	for (i = 0; i < RL_PHY_TIMEOUT; i++) {
465 		rval = CSR_READ_4(sc, RL_PHYAR);
466 		if (rval & RL_PHYAR_BUSY)
467 			break;
468 		DELAY(25);
469 	}
470 
471 	if (i == RL_PHY_TIMEOUT) {
472 		device_printf(sc->rl_dev, "PHY read failed\n");
473 		return (0);
474 	}
475 
476 	/*
477 	 * Controller requires a 20us delay to process next MDIO request.
478 	 */
479 	DELAY(20);
480 
481 	return (rval & RL_PHYAR_PHYDATA);
482 }
483 
484 static int
485 re_gmii_writereg(device_t dev, int phy, int reg, int data)
486 {
487 	struct rl_softc		*sc;
488 	u_int32_t		rval;
489 	int			i;
490 
491 	sc = device_get_softc(dev);
492 
493 	CSR_WRITE_4(sc, RL_PHYAR, (reg << 16) |
494 	    (data & RL_PHYAR_PHYDATA) | RL_PHYAR_BUSY);
495 
496 	for (i = 0; i < RL_PHY_TIMEOUT; i++) {
497 		rval = CSR_READ_4(sc, RL_PHYAR);
498 		if (!(rval & RL_PHYAR_BUSY))
499 			break;
500 		DELAY(25);
501 	}
502 
503 	if (i == RL_PHY_TIMEOUT) {
504 		device_printf(sc->rl_dev, "PHY write failed\n");
505 		return (0);
506 	}
507 
508 	/*
509 	 * Controller requires a 20us delay to process next MDIO request.
510 	 */
511 	DELAY(20);
512 
513 	return (0);
514 }
515 
516 static int
517 re_miibus_readreg(device_t dev, int phy, int reg)
518 {
519 	struct rl_softc		*sc;
520 	u_int16_t		rval = 0;
521 	u_int16_t		re8139_reg = 0;
522 
523 	sc = device_get_softc(dev);
524 
525 	if (sc->rl_type == RL_8169) {
526 		rval = re_gmii_readreg(dev, phy, reg);
527 		return (rval);
528 	}
529 
530 	switch (reg) {
531 	case MII_BMCR:
532 		re8139_reg = RL_BMCR;
533 		break;
534 	case MII_BMSR:
535 		re8139_reg = RL_BMSR;
536 		break;
537 	case MII_ANAR:
538 		re8139_reg = RL_ANAR;
539 		break;
540 	case MII_ANER:
541 		re8139_reg = RL_ANER;
542 		break;
543 	case MII_ANLPAR:
544 		re8139_reg = RL_LPAR;
545 		break;
546 	case MII_PHYIDR1:
547 	case MII_PHYIDR2:
548 		return (0);
549 	/*
550 	 * Allow the rlphy driver to read the media status
551 	 * register. If we have a link partner which does not
552 	 * support NWAY, this is the register which will tell
553 	 * us the results of parallel detection.
554 	 */
555 	case RL_MEDIASTAT:
556 		rval = CSR_READ_1(sc, RL_MEDIASTAT);
557 		return (rval);
558 	default:
559 		device_printf(sc->rl_dev, "bad phy register\n");
560 		return (0);
561 	}
562 	rval = CSR_READ_2(sc, re8139_reg);
563 	if (sc->rl_type == RL_8139CPLUS && re8139_reg == RL_BMCR) {
564 		/* 8139C+ has different bit layout. */
565 		rval &= ~(BMCR_LOOP | BMCR_ISO);
566 	}
567 	return (rval);
568 }
569 
570 static int
571 re_miibus_writereg(device_t dev, int phy, int reg, int data)
572 {
573 	struct rl_softc		*sc;
574 	u_int16_t		re8139_reg = 0;
575 	int			rval = 0;
576 
577 	sc = device_get_softc(dev);
578 
579 	if (sc->rl_type == RL_8169) {
580 		rval = re_gmii_writereg(dev, phy, reg, data);
581 		return (rval);
582 	}
583 
584 	switch (reg) {
585 	case MII_BMCR:
586 		re8139_reg = RL_BMCR;
587 		if (sc->rl_type == RL_8139CPLUS) {
588 			/* 8139C+ has different bit layout. */
589 			data &= ~(BMCR_LOOP | BMCR_ISO);
590 		}
591 		break;
592 	case MII_BMSR:
593 		re8139_reg = RL_BMSR;
594 		break;
595 	case MII_ANAR:
596 		re8139_reg = RL_ANAR;
597 		break;
598 	case MII_ANER:
599 		re8139_reg = RL_ANER;
600 		break;
601 	case MII_ANLPAR:
602 		re8139_reg = RL_LPAR;
603 		break;
604 	case MII_PHYIDR1:
605 	case MII_PHYIDR2:
606 		return (0);
607 		break;
608 	default:
609 		device_printf(sc->rl_dev, "bad phy register\n");
610 		return (0);
611 	}
612 	CSR_WRITE_2(sc, re8139_reg, data);
613 	return (0);
614 }
615 
616 static void
617 re_miibus_statchg(device_t dev)
618 {
619 	struct rl_softc		*sc;
620 	struct ifnet		*ifp;
621 	struct mii_data		*mii;
622 
623 	sc = device_get_softc(dev);
624 	mii = device_get_softc(sc->rl_miibus);
625 	ifp = sc->rl_ifp;
626 	if (mii == NULL || ifp == NULL ||
627 	    (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
628 		return;
629 
630 	sc->rl_flags &= ~RL_FLAG_LINK;
631 	if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
632 	    (IFM_ACTIVE | IFM_AVALID)) {
633 		switch (IFM_SUBTYPE(mii->mii_media_active)) {
634 		case IFM_10_T:
635 		case IFM_100_TX:
636 			sc->rl_flags |= RL_FLAG_LINK;
637 			break;
638 		case IFM_1000_T:
639 			if ((sc->rl_flags & RL_FLAG_FASTETHER) != 0)
640 				break;
641 			sc->rl_flags |= RL_FLAG_LINK;
642 			break;
643 		default:
644 			break;
645 		}
646 	}
647 	/*
648 	 * RealTek controllers do not provide any interface to the RX/TX
649 	 * MACs for resolved speed, duplex and flow-control parameters.
650 	 */
651 }
652 
653 /*
654  * Set the RX configuration and 64-bit multicast hash filter.
655  */
656 static void
657 re_set_rxmode(struct rl_softc *sc)
658 {
659 	struct ifnet		*ifp;
660 	struct ifmultiaddr	*ifma;
661 	uint32_t		hashes[2] = { 0, 0 };
662 	uint32_t		h, rxfilt;
663 
664 	RL_LOCK_ASSERT(sc);
665 
666 	ifp = sc->rl_ifp;
667 
668 	rxfilt = RL_RXCFG_CONFIG | RL_RXCFG_RX_INDIV | RL_RXCFG_RX_BROAD;
669 	if ((sc->rl_flags & RL_FLAG_EARLYOFF) != 0)
670 		rxfilt |= RL_RXCFG_EARLYOFF;
671 	else if ((sc->rl_flags & RL_FLAG_8168G_PLUS) != 0)
672 		rxfilt |= RL_RXCFG_EARLYOFFV2;
673 
674 	if (ifp->if_flags & (IFF_ALLMULTI | IFF_PROMISC)) {
675 		if (ifp->if_flags & IFF_PROMISC)
676 			rxfilt |= RL_RXCFG_RX_ALLPHYS;
677 		/*
678 		 * Unlike other hardwares, we have to explicitly set
679 		 * RL_RXCFG_RX_MULTI to receive multicast frames in
680 		 * promiscuous mode.
681 		 */
682 		rxfilt |= RL_RXCFG_RX_MULTI;
683 		hashes[0] = hashes[1] = 0xffffffff;
684 		goto done;
685 	}
686 
687 	if_maddr_rlock(ifp);
688 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
689 		if (ifma->ifma_addr->sa_family != AF_LINK)
690 			continue;
691 		h = ether_crc32_be(LLADDR((struct sockaddr_dl *)
692 		    ifma->ifma_addr), ETHER_ADDR_LEN) >> 26;
693 		if (h < 32)
694 			hashes[0] |= (1 << h);
695 		else
696 			hashes[1] |= (1 << (h - 32));
697 	}
698 	if_maddr_runlock(ifp);
699 
700 	if (hashes[0] != 0 || hashes[1] != 0) {
701 		/*
702 		 * For some unfathomable reason, RealTek decided to
703 		 * reverse the order of the multicast hash registers
704 		 * in the PCI Express parts.  This means we have to
705 		 * write the hash pattern in reverse order for those
706 		 * devices.
707 		 */
708 		if ((sc->rl_flags & RL_FLAG_PCIE) != 0) {
709 			h = bswap32(hashes[0]);
710 			hashes[0] = bswap32(hashes[1]);
711 			hashes[1] = h;
712 		}
713 		rxfilt |= RL_RXCFG_RX_MULTI;
714 	}
715 
716 	if  (sc->rl_hwrev->rl_rev == RL_HWREV_8168F) {
717 		/* Disable multicast filtering due to silicon bug. */
718 		hashes[0] = 0xffffffff;
719 		hashes[1] = 0xffffffff;
720 	}
721 
722 done:
723 	CSR_WRITE_4(sc, RL_MAR0, hashes[0]);
724 	CSR_WRITE_4(sc, RL_MAR4, hashes[1]);
725 	CSR_WRITE_4(sc, RL_RXCFG, rxfilt);
726 }
727 
728 static void
729 re_reset(struct rl_softc *sc)
730 {
731 	int			i;
732 
733 	RL_LOCK_ASSERT(sc);
734 
735 	CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_RESET);
736 
737 	for (i = 0; i < RL_TIMEOUT; i++) {
738 		DELAY(10);
739 		if (!(CSR_READ_1(sc, RL_COMMAND) & RL_CMD_RESET))
740 			break;
741 	}
742 	if (i == RL_TIMEOUT)
743 		device_printf(sc->rl_dev, "reset never completed!\n");
744 
745 	if ((sc->rl_flags & RL_FLAG_MACRESET) != 0)
746 		CSR_WRITE_1(sc, 0x82, 1);
747 	if (sc->rl_hwrev->rl_rev == RL_HWREV_8169S)
748 		re_gmii_writereg(sc->rl_dev, 1, 0x0b, 0);
749 }
750 
751 #ifdef RE_DIAG
752 
753 /*
754  * The following routine is designed to test for a defect on some
755  * 32-bit 8169 cards. Some of these NICs have the REQ64# and ACK64#
756  * lines connected to the bus, however for a 32-bit only card, they
757  * should be pulled high. The result of this defect is that the
758  * NIC will not work right if you plug it into a 64-bit slot: DMA
759  * operations will be done with 64-bit transfers, which will fail
760  * because the 64-bit data lines aren't connected.
761  *
762  * There's no way to work around this (short of talking a soldering
763  * iron to the board), however we can detect it. The method we use
764  * here is to put the NIC into digital loopback mode, set the receiver
765  * to promiscuous mode, and then try to send a frame. We then compare
766  * the frame data we sent to what was received. If the data matches,
767  * then the NIC is working correctly, otherwise we know the user has
768  * a defective NIC which has been mistakenly plugged into a 64-bit PCI
769  * slot. In the latter case, there's no way the NIC can work correctly,
770  * so we print out a message on the console and abort the device attach.
771  */
772 
773 static int
774 re_diag(struct rl_softc *sc)
775 {
776 	struct ifnet		*ifp = sc->rl_ifp;
777 	struct mbuf		*m0;
778 	struct ether_header	*eh;
779 	struct rl_desc		*cur_rx;
780 	u_int16_t		status;
781 	u_int32_t		rxstat;
782 	int			total_len, i, error = 0, phyaddr;
783 	u_int8_t		dst[] = { 0x00, 'h', 'e', 'l', 'l', 'o' };
784 	u_int8_t		src[] = { 0x00, 'w', 'o', 'r', 'l', 'd' };
785 
786 	/* Allocate a single mbuf */
787 	MGETHDR(m0, M_NOWAIT, MT_DATA);
788 	if (m0 == NULL)
789 		return (ENOBUFS);
790 
791 	RL_LOCK(sc);
792 
793 	/*
794 	 * Initialize the NIC in test mode. This sets the chip up
795 	 * so that it can send and receive frames, but performs the
796 	 * following special functions:
797 	 * - Puts receiver in promiscuous mode
798 	 * - Enables digital loopback mode
799 	 * - Leaves interrupts turned off
800 	 */
801 
802 	ifp->if_flags |= IFF_PROMISC;
803 	sc->rl_testmode = 1;
804 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
805 	re_init_locked(sc);
806 	sc->rl_flags |= RL_FLAG_LINK;
807 	if (sc->rl_type == RL_8169)
808 		phyaddr = 1;
809 	else
810 		phyaddr = 0;
811 
812 	re_miibus_writereg(sc->rl_dev, phyaddr, MII_BMCR, BMCR_RESET);
813 	for (i = 0; i < RL_TIMEOUT; i++) {
814 		status = re_miibus_readreg(sc->rl_dev, phyaddr, MII_BMCR);
815 		if (!(status & BMCR_RESET))
816 			break;
817 	}
818 
819 	re_miibus_writereg(sc->rl_dev, phyaddr, MII_BMCR, BMCR_LOOP);
820 	CSR_WRITE_2(sc, RL_ISR, RL_INTRS);
821 
822 	DELAY(100000);
823 
824 	/* Put some data in the mbuf */
825 
826 	eh = mtod(m0, struct ether_header *);
827 	bcopy ((char *)&dst, eh->ether_dhost, ETHER_ADDR_LEN);
828 	bcopy ((char *)&src, eh->ether_shost, ETHER_ADDR_LEN);
829 	eh->ether_type = htons(ETHERTYPE_IP);
830 	m0->m_pkthdr.len = m0->m_len = ETHER_MIN_LEN - ETHER_CRC_LEN;
831 
832 	/*
833 	 * Queue the packet, start transmission.
834 	 * Note: IF_HANDOFF() ultimately calls re_start() for us.
835 	 */
836 
837 	CSR_WRITE_2(sc, RL_ISR, 0xFFFF);
838 	RL_UNLOCK(sc);
839 	/* XXX: re_diag must not be called when in ALTQ mode */
840 	IF_HANDOFF(&ifp->if_snd, m0, ifp);
841 	RL_LOCK(sc);
842 	m0 = NULL;
843 
844 	/* Wait for it to propagate through the chip */
845 
846 	DELAY(100000);
847 	for (i = 0; i < RL_TIMEOUT; i++) {
848 		status = CSR_READ_2(sc, RL_ISR);
849 		CSR_WRITE_2(sc, RL_ISR, status);
850 		if ((status & (RL_ISR_TIMEOUT_EXPIRED|RL_ISR_RX_OK)) ==
851 		    (RL_ISR_TIMEOUT_EXPIRED|RL_ISR_RX_OK))
852 			break;
853 		DELAY(10);
854 	}
855 
856 	if (i == RL_TIMEOUT) {
857 		device_printf(sc->rl_dev,
858 		    "diagnostic failed, failed to receive packet in"
859 		    " loopback mode\n");
860 		error = EIO;
861 		goto done;
862 	}
863 
864 	/*
865 	 * The packet should have been dumped into the first
866 	 * entry in the RX DMA ring. Grab it from there.
867 	 */
868 
869 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
870 	    sc->rl_ldata.rl_rx_list_map,
871 	    BUS_DMASYNC_POSTREAD);
872 	bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag,
873 	    sc->rl_ldata.rl_rx_desc[0].rx_dmamap,
874 	    BUS_DMASYNC_POSTREAD);
875 	bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag,
876 	    sc->rl_ldata.rl_rx_desc[0].rx_dmamap);
877 
878 	m0 = sc->rl_ldata.rl_rx_desc[0].rx_m;
879 	sc->rl_ldata.rl_rx_desc[0].rx_m = NULL;
880 	eh = mtod(m0, struct ether_header *);
881 
882 	cur_rx = &sc->rl_ldata.rl_rx_list[0];
883 	total_len = RL_RXBYTES(cur_rx);
884 	rxstat = le32toh(cur_rx->rl_cmdstat);
885 
886 	if (total_len != ETHER_MIN_LEN) {
887 		device_printf(sc->rl_dev,
888 		    "diagnostic failed, received short packet\n");
889 		error = EIO;
890 		goto done;
891 	}
892 
893 	/* Test that the received packet data matches what we sent. */
894 
895 	if (bcmp((char *)&eh->ether_dhost, (char *)&dst, ETHER_ADDR_LEN) ||
896 	    bcmp((char *)&eh->ether_shost, (char *)&src, ETHER_ADDR_LEN) ||
897 	    ntohs(eh->ether_type) != ETHERTYPE_IP) {
898 		device_printf(sc->rl_dev, "WARNING, DMA FAILURE!\n");
899 		device_printf(sc->rl_dev, "expected TX data: %6D/%6D/0x%x\n",
900 		    dst, ":", src, ":", ETHERTYPE_IP);
901 		device_printf(sc->rl_dev, "received RX data: %6D/%6D/0x%x\n",
902 		    eh->ether_dhost, ":", eh->ether_shost, ":",
903 		    ntohs(eh->ether_type));
904 		device_printf(sc->rl_dev, "You may have a defective 32-bit "
905 		    "NIC plugged into a 64-bit PCI slot.\n");
906 		device_printf(sc->rl_dev, "Please re-install the NIC in a "
907 		    "32-bit slot for proper operation.\n");
908 		device_printf(sc->rl_dev, "Read the re(4) man page for more "
909 		    "details.\n");
910 		error = EIO;
911 	}
912 
913 done:
914 	/* Turn interface off, release resources */
915 
916 	sc->rl_testmode = 0;
917 	sc->rl_flags &= ~RL_FLAG_LINK;
918 	ifp->if_flags &= ~IFF_PROMISC;
919 	re_stop(sc);
920 	if (m0 != NULL)
921 		m_freem(m0);
922 
923 	RL_UNLOCK(sc);
924 
925 	return (error);
926 }
927 
928 #endif
929 
930 /*
931  * Probe for a RealTek 8139C+/8169/8110 chip. Check the PCI vendor and device
932  * IDs against our list and return a device name if we find a match.
933  */
934 static int
935 re_probe(device_t dev)
936 {
937 	const struct rl_type	*t;
938 	uint16_t		devid, vendor;
939 	uint16_t		revid, sdevid;
940 	int			i;
941 
942 	vendor = pci_get_vendor(dev);
943 	devid = pci_get_device(dev);
944 	revid = pci_get_revid(dev);
945 	sdevid = pci_get_subdevice(dev);
946 
947 	if (vendor == LINKSYS_VENDORID && devid == LINKSYS_DEVICEID_EG1032) {
948 		if (sdevid != LINKSYS_SUBDEVICE_EG1032_REV3) {
949 			/*
950 			 * Only attach to rev. 3 of the Linksys EG1032 adapter.
951 			 * Rev. 2 is supported by sk(4).
952 			 */
953 			return (ENXIO);
954 		}
955 	}
956 
957 	if (vendor == RT_VENDORID && devid == RT_DEVICEID_8139) {
958 		if (revid != 0x20) {
959 			/* 8139, let rl(4) take care of this device. */
960 			return (ENXIO);
961 		}
962 	}
963 
964 	t = re_devs;
965 	for (i = 0; i < nitems(re_devs); i++, t++) {
966 		if (vendor == t->rl_vid && devid == t->rl_did) {
967 			device_set_desc(dev, t->rl_name);
968 			return (BUS_PROBE_DEFAULT);
969 		}
970 	}
971 
972 	return (ENXIO);
973 }
974 
975 /*
976  * Map a single buffer address.
977  */
978 
979 static void
980 re_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
981 {
982 	bus_addr_t		*addr;
983 
984 	if (error)
985 		return;
986 
987 	KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg));
988 	addr = arg;
989 	*addr = segs->ds_addr;
990 }
991 
992 static int
993 re_allocmem(device_t dev, struct rl_softc *sc)
994 {
995 	bus_addr_t		lowaddr;
996 	bus_size_t		rx_list_size, tx_list_size;
997 	int			error;
998 	int			i;
999 
1000 	rx_list_size = sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc);
1001 	tx_list_size = sc->rl_ldata.rl_tx_desc_cnt * sizeof(struct rl_desc);
1002 
1003 	/*
1004 	 * Allocate the parent bus DMA tag appropriate for PCI.
1005 	 * In order to use DAC, RL_CPLUSCMD_PCI_DAC bit of RL_CPLUS_CMD
1006 	 * register should be set. However some RealTek chips are known
1007 	 * to be buggy on DAC handling, therefore disable DAC by limiting
1008 	 * DMA address space to 32bit. PCIe variants of RealTek chips
1009 	 * may not have the limitation.
1010 	 */
1011 	lowaddr = BUS_SPACE_MAXADDR;
1012 	if ((sc->rl_flags & RL_FLAG_PCIE) == 0)
1013 		lowaddr = BUS_SPACE_MAXADDR_32BIT;
1014 	error = bus_dma_tag_create(bus_get_dma_tag(dev), 1, 0,
1015 	    lowaddr, BUS_SPACE_MAXADDR, NULL, NULL,
1016 	    BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0,
1017 	    NULL, NULL, &sc->rl_parent_tag);
1018 	if (error) {
1019 		device_printf(dev, "could not allocate parent DMA tag\n");
1020 		return (error);
1021 	}
1022 
1023 	/*
1024 	 * Allocate map for TX mbufs.
1025 	 */
1026 	error = bus_dma_tag_create(sc->rl_parent_tag, 1, 0,
1027 	    BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
1028 	    NULL, MCLBYTES * RL_NTXSEGS, RL_NTXSEGS, 4096, 0,
1029 	    NULL, NULL, &sc->rl_ldata.rl_tx_mtag);
1030 	if (error) {
1031 		device_printf(dev, "could not allocate TX DMA tag\n");
1032 		return (error);
1033 	}
1034 
1035 	/*
1036 	 * Allocate map for RX mbufs.
1037 	 */
1038 
1039 	if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) {
1040 		error = bus_dma_tag_create(sc->rl_parent_tag, sizeof(uint64_t),
1041 		    0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
1042 		    MJUM9BYTES, 1, MJUM9BYTES, 0, NULL, NULL,
1043 		    &sc->rl_ldata.rl_jrx_mtag);
1044 		if (error) {
1045 			device_printf(dev,
1046 			    "could not allocate jumbo RX DMA tag\n");
1047 			return (error);
1048 		}
1049 	}
1050 	error = bus_dma_tag_create(sc->rl_parent_tag, sizeof(uint64_t), 0,
1051 	    BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
1052 	    MCLBYTES, 1, MCLBYTES, 0, NULL, NULL, &sc->rl_ldata.rl_rx_mtag);
1053 	if (error) {
1054 		device_printf(dev, "could not allocate RX DMA tag\n");
1055 		return (error);
1056 	}
1057 
1058 	/*
1059 	 * Allocate map for TX descriptor list.
1060 	 */
1061 	error = bus_dma_tag_create(sc->rl_parent_tag, RL_RING_ALIGN,
1062 	    0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL,
1063 	    NULL, tx_list_size, 1, tx_list_size, 0,
1064 	    NULL, NULL, &sc->rl_ldata.rl_tx_list_tag);
1065 	if (error) {
1066 		device_printf(dev, "could not allocate TX DMA ring tag\n");
1067 		return (error);
1068 	}
1069 
1070 	/* Allocate DMA'able memory for the TX ring */
1071 
1072 	error = bus_dmamem_alloc(sc->rl_ldata.rl_tx_list_tag,
1073 	    (void **)&sc->rl_ldata.rl_tx_list,
1074 	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
1075 	    &sc->rl_ldata.rl_tx_list_map);
1076 	if (error) {
1077 		device_printf(dev, "could not allocate TX DMA ring\n");
1078 		return (error);
1079 	}
1080 
1081 	/* Load the map for the TX ring. */
1082 
1083 	sc->rl_ldata.rl_tx_list_addr = 0;
1084 	error = bus_dmamap_load(sc->rl_ldata.rl_tx_list_tag,
1085 	     sc->rl_ldata.rl_tx_list_map, sc->rl_ldata.rl_tx_list,
1086 	     tx_list_size, re_dma_map_addr,
1087 	     &sc->rl_ldata.rl_tx_list_addr, BUS_DMA_NOWAIT);
1088 	if (error != 0 || sc->rl_ldata.rl_tx_list_addr == 0) {
1089 		device_printf(dev, "could not load TX DMA ring\n");
1090 		return (ENOMEM);
1091 	}
1092 
1093 	/* Create DMA maps for TX buffers */
1094 
1095 	for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) {
1096 		error = bus_dmamap_create(sc->rl_ldata.rl_tx_mtag, 0,
1097 		    &sc->rl_ldata.rl_tx_desc[i].tx_dmamap);
1098 		if (error) {
1099 			device_printf(dev, "could not create DMA map for TX\n");
1100 			return (error);
1101 		}
1102 	}
1103 
1104 	/*
1105 	 * Allocate map for RX descriptor list.
1106 	 */
1107 	error = bus_dma_tag_create(sc->rl_parent_tag, RL_RING_ALIGN,
1108 	    0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL,
1109 	    NULL, rx_list_size, 1, rx_list_size, 0,
1110 	    NULL, NULL, &sc->rl_ldata.rl_rx_list_tag);
1111 	if (error) {
1112 		device_printf(dev, "could not create RX DMA ring tag\n");
1113 		return (error);
1114 	}
1115 
1116 	/* Allocate DMA'able memory for the RX ring */
1117 
1118 	error = bus_dmamem_alloc(sc->rl_ldata.rl_rx_list_tag,
1119 	    (void **)&sc->rl_ldata.rl_rx_list,
1120 	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
1121 	    &sc->rl_ldata.rl_rx_list_map);
1122 	if (error) {
1123 		device_printf(dev, "could not allocate RX DMA ring\n");
1124 		return (error);
1125 	}
1126 
1127 	/* Load the map for the RX ring. */
1128 
1129 	sc->rl_ldata.rl_rx_list_addr = 0;
1130 	error = bus_dmamap_load(sc->rl_ldata.rl_rx_list_tag,
1131 	     sc->rl_ldata.rl_rx_list_map, sc->rl_ldata.rl_rx_list,
1132 	     rx_list_size, re_dma_map_addr,
1133 	     &sc->rl_ldata.rl_rx_list_addr, BUS_DMA_NOWAIT);
1134 	if (error != 0 || sc->rl_ldata.rl_rx_list_addr == 0) {
1135 		device_printf(dev, "could not load RX DMA ring\n");
1136 		return (ENOMEM);
1137 	}
1138 
1139 	/* Create DMA maps for RX buffers */
1140 
1141 	if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) {
1142 		error = bus_dmamap_create(sc->rl_ldata.rl_jrx_mtag, 0,
1143 		    &sc->rl_ldata.rl_jrx_sparemap);
1144 		if (error) {
1145 			device_printf(dev,
1146 			    "could not create spare DMA map for jumbo RX\n");
1147 			return (error);
1148 		}
1149 		for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
1150 			error = bus_dmamap_create(sc->rl_ldata.rl_jrx_mtag, 0,
1151 			    &sc->rl_ldata.rl_jrx_desc[i].rx_dmamap);
1152 			if (error) {
1153 				device_printf(dev,
1154 				    "could not create DMA map for jumbo RX\n");
1155 				return (error);
1156 			}
1157 		}
1158 	}
1159 	error = bus_dmamap_create(sc->rl_ldata.rl_rx_mtag, 0,
1160 	    &sc->rl_ldata.rl_rx_sparemap);
1161 	if (error) {
1162 		device_printf(dev, "could not create spare DMA map for RX\n");
1163 		return (error);
1164 	}
1165 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
1166 		error = bus_dmamap_create(sc->rl_ldata.rl_rx_mtag, 0,
1167 		    &sc->rl_ldata.rl_rx_desc[i].rx_dmamap);
1168 		if (error) {
1169 			device_printf(dev, "could not create DMA map for RX\n");
1170 			return (error);
1171 		}
1172 	}
1173 
1174 	/* Create DMA map for statistics. */
1175 	error = bus_dma_tag_create(sc->rl_parent_tag, RL_DUMP_ALIGN, 0,
1176 	    BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
1177 	    sizeof(struct rl_stats), 1, sizeof(struct rl_stats), 0, NULL, NULL,
1178 	    &sc->rl_ldata.rl_stag);
1179 	if (error) {
1180 		device_printf(dev, "could not create statistics DMA tag\n");
1181 		return (error);
1182 	}
1183 	/* Allocate DMA'able memory for statistics. */
1184 	error = bus_dmamem_alloc(sc->rl_ldata.rl_stag,
1185 	    (void **)&sc->rl_ldata.rl_stats,
1186 	    BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO,
1187 	    &sc->rl_ldata.rl_smap);
1188 	if (error) {
1189 		device_printf(dev,
1190 		    "could not allocate statistics DMA memory\n");
1191 		return (error);
1192 	}
1193 	/* Load the map for statistics. */
1194 	sc->rl_ldata.rl_stats_addr = 0;
1195 	error = bus_dmamap_load(sc->rl_ldata.rl_stag, sc->rl_ldata.rl_smap,
1196 	    sc->rl_ldata.rl_stats, sizeof(struct rl_stats), re_dma_map_addr,
1197 	     &sc->rl_ldata.rl_stats_addr, BUS_DMA_NOWAIT);
1198 	if (error != 0 || sc->rl_ldata.rl_stats_addr == 0) {
1199 		device_printf(dev, "could not load statistics DMA memory\n");
1200 		return (ENOMEM);
1201 	}
1202 
1203 	return (0);
1204 }
1205 
1206 /*
1207  * Attach the interface. Allocate softc structures, do ifmedia
1208  * setup and ethernet/BPF attach.
1209  */
1210 static int
1211 re_attach(device_t dev)
1212 {
1213 	u_char			eaddr[ETHER_ADDR_LEN];
1214 	u_int16_t		as[ETHER_ADDR_LEN / 2];
1215 	struct rl_softc		*sc;
1216 	struct ifnet		*ifp;
1217 	const struct rl_hwrev	*hw_rev;
1218 	int			capmask, error = 0, hwrev, i, msic, msixc,
1219 				phy, reg, rid;
1220 	u_int32_t		cap, ctl;
1221 	u_int16_t		devid, re_did = 0;
1222 	uint8_t			cfg;
1223 
1224 	sc = device_get_softc(dev);
1225 	sc->rl_dev = dev;
1226 
1227 	mtx_init(&sc->rl_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
1228 	    MTX_DEF);
1229 	callout_init_mtx(&sc->rl_stat_callout, &sc->rl_mtx, 0);
1230 
1231 	/*
1232 	 * Map control/status registers.
1233 	 */
1234 	pci_enable_busmaster(dev);
1235 
1236 	devid = pci_get_device(dev);
1237 	/*
1238 	 * Prefer memory space register mapping over IO space.
1239 	 * Because RTL8169SC does not seem to work when memory mapping
1240 	 * is used always activate io mapping.
1241 	 */
1242 	if (devid == RT_DEVICEID_8169SC)
1243 		prefer_iomap = 1;
1244 	if (prefer_iomap == 0) {
1245 		sc->rl_res_id = PCIR_BAR(1);
1246 		sc->rl_res_type = SYS_RES_MEMORY;
1247 		/* RTL8168/8101E seems to use different BARs. */
1248 		if (devid == RT_DEVICEID_8168 || devid == RT_DEVICEID_8101E)
1249 			sc->rl_res_id = PCIR_BAR(2);
1250 	} else {
1251 		sc->rl_res_id = PCIR_BAR(0);
1252 		sc->rl_res_type = SYS_RES_IOPORT;
1253 	}
1254 	sc->rl_res = bus_alloc_resource_any(dev, sc->rl_res_type,
1255 	    &sc->rl_res_id, RF_ACTIVE);
1256 	if (sc->rl_res == NULL && prefer_iomap == 0) {
1257 		sc->rl_res_id = PCIR_BAR(0);
1258 		sc->rl_res_type = SYS_RES_IOPORT;
1259 		sc->rl_res = bus_alloc_resource_any(dev, sc->rl_res_type,
1260 		    &sc->rl_res_id, RF_ACTIVE);
1261 	}
1262 	if (sc->rl_res == NULL) {
1263 		device_printf(dev, "couldn't map ports/memory\n");
1264 		error = ENXIO;
1265 		goto fail;
1266 	}
1267 
1268 	sc->rl_btag = rman_get_bustag(sc->rl_res);
1269 	sc->rl_bhandle = rman_get_bushandle(sc->rl_res);
1270 
1271 	msic = pci_msi_count(dev);
1272 	msixc = pci_msix_count(dev);
1273 	if (pci_find_cap(dev, PCIY_EXPRESS, &reg) == 0) {
1274 		sc->rl_flags |= RL_FLAG_PCIE;
1275 		sc->rl_expcap = reg;
1276 	}
1277 	if (bootverbose) {
1278 		device_printf(dev, "MSI count : %d\n", msic);
1279 		device_printf(dev, "MSI-X count : %d\n", msixc);
1280 	}
1281 	if (msix_disable > 0)
1282 		msixc = 0;
1283 	if (msi_disable > 0)
1284 		msic = 0;
1285 	/* Prefer MSI-X to MSI. */
1286 	if (msixc > 0) {
1287 		msixc = RL_MSI_MESSAGES;
1288 		rid = PCIR_BAR(4);
1289 		sc->rl_res_pba = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1290 		    &rid, RF_ACTIVE);
1291 		if (sc->rl_res_pba == NULL) {
1292 			device_printf(sc->rl_dev,
1293 			    "could not allocate MSI-X PBA resource\n");
1294 		}
1295 		if (sc->rl_res_pba != NULL &&
1296 		    pci_alloc_msix(dev, &msixc) == 0) {
1297 			if (msixc == RL_MSI_MESSAGES) {
1298 				device_printf(dev, "Using %d MSI-X message\n",
1299 				    msixc);
1300 				sc->rl_flags |= RL_FLAG_MSIX;
1301 			} else
1302 				pci_release_msi(dev);
1303 		}
1304 		if ((sc->rl_flags & RL_FLAG_MSIX) == 0) {
1305 			if (sc->rl_res_pba != NULL)
1306 				bus_release_resource(dev, SYS_RES_MEMORY, rid,
1307 				    sc->rl_res_pba);
1308 			sc->rl_res_pba = NULL;
1309 			msixc = 0;
1310 		}
1311 	}
1312 	/* Prefer MSI to INTx. */
1313 	if (msixc == 0 && msic > 0) {
1314 		msic = RL_MSI_MESSAGES;
1315 		if (pci_alloc_msi(dev, &msic) == 0) {
1316 			if (msic == RL_MSI_MESSAGES) {
1317 				device_printf(dev, "Using %d MSI message\n",
1318 				    msic);
1319 				sc->rl_flags |= RL_FLAG_MSI;
1320 				/* Explicitly set MSI enable bit. */
1321 				CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
1322 				cfg = CSR_READ_1(sc, RL_CFG2);
1323 				cfg |= RL_CFG2_MSI;
1324 				CSR_WRITE_1(sc, RL_CFG2, cfg);
1325 				CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
1326 			} else
1327 				pci_release_msi(dev);
1328 		}
1329 		if ((sc->rl_flags & RL_FLAG_MSI) == 0)
1330 			msic = 0;
1331 	}
1332 
1333 	/* Allocate interrupt */
1334 	if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) == 0) {
1335 		rid = 0;
1336 		sc->rl_irq[0] = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
1337 		    RF_SHAREABLE | RF_ACTIVE);
1338 		if (sc->rl_irq[0] == NULL) {
1339 			device_printf(dev, "couldn't allocate IRQ resources\n");
1340 			error = ENXIO;
1341 			goto fail;
1342 		}
1343 	} else {
1344 		for (i = 0, rid = 1; i < RL_MSI_MESSAGES; i++, rid++) {
1345 			sc->rl_irq[i] = bus_alloc_resource_any(dev,
1346 			    SYS_RES_IRQ, &rid, RF_ACTIVE);
1347 			if (sc->rl_irq[i] == NULL) {
1348 				device_printf(dev,
1349 				    "couldn't allocate IRQ resources for "
1350 				    "message %d\n", rid);
1351 				error = ENXIO;
1352 				goto fail;
1353 			}
1354 		}
1355 	}
1356 
1357 	if ((sc->rl_flags & RL_FLAG_MSI) == 0) {
1358 		CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
1359 		cfg = CSR_READ_1(sc, RL_CFG2);
1360 		if ((cfg & RL_CFG2_MSI) != 0) {
1361 			device_printf(dev, "turning off MSI enable bit.\n");
1362 			cfg &= ~RL_CFG2_MSI;
1363 			CSR_WRITE_1(sc, RL_CFG2, cfg);
1364 		}
1365 		CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
1366 	}
1367 
1368 	/* Disable ASPM L0S/L1 and CLKREQ. */
1369 	if (sc->rl_expcap != 0) {
1370 		cap = pci_read_config(dev, sc->rl_expcap +
1371 		    PCIER_LINK_CAP, 2);
1372 		if ((cap & PCIEM_LINK_CAP_ASPM) != 0) {
1373 			ctl = pci_read_config(dev, sc->rl_expcap +
1374 			    PCIER_LINK_CTL, 2);
1375 			if ((ctl & (PCIEM_LINK_CTL_ECPM |
1376 			    PCIEM_LINK_CTL_ASPMC))!= 0) {
1377 				ctl &= ~(PCIEM_LINK_CTL_ECPM |
1378 				    PCIEM_LINK_CTL_ASPMC);
1379 				pci_write_config(dev, sc->rl_expcap +
1380 				    PCIER_LINK_CTL, ctl, 2);
1381 				device_printf(dev, "ASPM disabled\n");
1382 			}
1383 		} else
1384 			device_printf(dev, "no ASPM capability\n");
1385 	}
1386 
1387 	hw_rev = re_hwrevs;
1388 	hwrev = CSR_READ_4(sc, RL_TXCFG);
1389 	switch (hwrev & 0x70000000) {
1390 	case 0x00000000:
1391 	case 0x10000000:
1392 		device_printf(dev, "Chip rev. 0x%08x\n", hwrev & 0xfc800000);
1393 		hwrev &= (RL_TXCFG_HWREV | 0x80000000);
1394 		break;
1395 	default:
1396 		device_printf(dev, "Chip rev. 0x%08x\n", hwrev & 0x7c800000);
1397 		sc->rl_macrev = hwrev & 0x00700000;
1398 		hwrev &= RL_TXCFG_HWREV;
1399 		break;
1400 	}
1401 	device_printf(dev, "MAC rev. 0x%08x\n", sc->rl_macrev);
1402 	while (hw_rev->rl_desc != NULL) {
1403 		if (hw_rev->rl_rev == hwrev) {
1404 			sc->rl_type = hw_rev->rl_type;
1405 			sc->rl_hwrev = hw_rev;
1406 			break;
1407 		}
1408 		hw_rev++;
1409 	}
1410 	if (hw_rev->rl_desc == NULL) {
1411 		device_printf(dev, "Unknown H/W revision: 0x%08x\n", hwrev);
1412 		error = ENXIO;
1413 		goto fail;
1414 	}
1415 
1416 	switch (hw_rev->rl_rev) {
1417 	case RL_HWREV_8139CPLUS:
1418 		sc->rl_flags |= RL_FLAG_FASTETHER | RL_FLAG_AUTOPAD;
1419 		break;
1420 	case RL_HWREV_8100E:
1421 	case RL_HWREV_8101E:
1422 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_FASTETHER;
1423 		break;
1424 	case RL_HWREV_8102E:
1425 	case RL_HWREV_8102EL:
1426 	case RL_HWREV_8102EL_SPIN1:
1427 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR | RL_FLAG_DESCV2 |
1428 		    RL_FLAG_MACSTAT | RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP |
1429 		    RL_FLAG_AUTOPAD;
1430 		break;
1431 	case RL_HWREV_8103E:
1432 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR | RL_FLAG_DESCV2 |
1433 		    RL_FLAG_MACSTAT | RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP |
1434 		    RL_FLAG_AUTOPAD | RL_FLAG_MACSLEEP;
1435 		break;
1436 	case RL_HWREV_8401E:
1437 	case RL_HWREV_8105E:
1438 	case RL_HWREV_8105E_SPIN1:
1439 	case RL_HWREV_8106E:
1440 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PHYWAKE_PM |
1441 		    RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT |
1442 		    RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP | RL_FLAG_AUTOPAD;
1443 		break;
1444 	case RL_HWREV_8402:
1445 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PHYWAKE_PM |
1446 		    RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT |
1447 		    RL_FLAG_FASTETHER | RL_FLAG_CMDSTOP | RL_FLAG_AUTOPAD |
1448 		    RL_FLAG_CMDSTOP_WAIT_TXQ;
1449 		break;
1450 	case RL_HWREV_8168B_SPIN1:
1451 	case RL_HWREV_8168B_SPIN2:
1452 		sc->rl_flags |= RL_FLAG_WOLRXENB;
1453 		/* FALLTHROUGH */
1454 	case RL_HWREV_8168B_SPIN3:
1455 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_MACSTAT;
1456 		break;
1457 	case RL_HWREV_8168C_SPIN2:
1458 		sc->rl_flags |= RL_FLAG_MACSLEEP;
1459 		/* FALLTHROUGH */
1460 	case RL_HWREV_8168C:
1461 		if (sc->rl_macrev == 0x00200000)
1462 			sc->rl_flags |= RL_FLAG_MACSLEEP;
1463 		/* FALLTHROUGH */
1464 	case RL_HWREV_8168CP:
1465 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR |
1466 		    RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP |
1467 		    RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2 | RL_FLAG_WOL_MANLINK;
1468 		break;
1469 	case RL_HWREV_8168D:
1470 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PHYWAKE_PM |
1471 		    RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT |
1472 		    RL_FLAG_CMDSTOP | RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2 |
1473 		    RL_FLAG_WOL_MANLINK;
1474 		break;
1475 	case RL_HWREV_8168DP:
1476 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR |
1477 		    RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_AUTOPAD |
1478 		    RL_FLAG_JUMBOV2 | RL_FLAG_WAIT_TXPOLL | RL_FLAG_WOL_MANLINK;
1479 		break;
1480 	case RL_HWREV_8168E:
1481 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PHYWAKE_PM |
1482 		    RL_FLAG_PAR | RL_FLAG_DESCV2 | RL_FLAG_MACSTAT |
1483 		    RL_FLAG_CMDSTOP | RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2 |
1484 		    RL_FLAG_WOL_MANLINK;
1485 		break;
1486 	case RL_HWREV_8168E_VL:
1487 	case RL_HWREV_8168F:
1488 		sc->rl_flags |= RL_FLAG_EARLYOFF;
1489 		/* FALLTHROUGH */
1490 	case RL_HWREV_8411:
1491 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR |
1492 		    RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP |
1493 		    RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2 |
1494 		    RL_FLAG_CMDSTOP_WAIT_TXQ | RL_FLAG_WOL_MANLINK;
1495 		break;
1496 	case RL_HWREV_8168EP:
1497 	case RL_HWREV_8168G:
1498 	case RL_HWREV_8411B:
1499 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR |
1500 		    RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP |
1501 		    RL_FLAG_AUTOPAD | RL_FLAG_JUMBOV2 |
1502 		    RL_FLAG_CMDSTOP_WAIT_TXQ | RL_FLAG_WOL_MANLINK |
1503 		    RL_FLAG_8168G_PLUS;
1504 		break;
1505 	case RL_HWREV_8168GU:
1506 	case RL_HWREV_8168H:
1507 		if (pci_get_device(dev) == RT_DEVICEID_8101E) {
1508 			/* RTL8106E(US), RTL8107E */
1509 			sc->rl_flags |= RL_FLAG_FASTETHER;
1510 		} else
1511 			sc->rl_flags |= RL_FLAG_JUMBOV2 | RL_FLAG_WOL_MANLINK;
1512 
1513 		sc->rl_flags |= RL_FLAG_PHYWAKE | RL_FLAG_PAR |
1514 		    RL_FLAG_DESCV2 | RL_FLAG_MACSTAT | RL_FLAG_CMDSTOP |
1515 		    RL_FLAG_AUTOPAD | RL_FLAG_CMDSTOP_WAIT_TXQ |
1516 		    RL_FLAG_8168G_PLUS;
1517 		break;
1518 	case RL_HWREV_8169_8110SB:
1519 	case RL_HWREV_8169_8110SBL:
1520 	case RL_HWREV_8169_8110SC:
1521 	case RL_HWREV_8169_8110SCE:
1522 		sc->rl_flags |= RL_FLAG_PHYWAKE;
1523 		/* FALLTHROUGH */
1524 	case RL_HWREV_8169:
1525 	case RL_HWREV_8169S:
1526 	case RL_HWREV_8110S:
1527 		sc->rl_flags |= RL_FLAG_MACRESET;
1528 		break;
1529 	default:
1530 		break;
1531 	}
1532 
1533 	if (sc->rl_hwrev->rl_rev == RL_HWREV_8139CPLUS) {
1534 		sc->rl_cfg0 = RL_8139_CFG0;
1535 		sc->rl_cfg1 = RL_8139_CFG1;
1536 		sc->rl_cfg2 = 0;
1537 		sc->rl_cfg3 = RL_8139_CFG3;
1538 		sc->rl_cfg4 = RL_8139_CFG4;
1539 		sc->rl_cfg5 = RL_8139_CFG5;
1540 	} else {
1541 		sc->rl_cfg0 = RL_CFG0;
1542 		sc->rl_cfg1 = RL_CFG1;
1543 		sc->rl_cfg2 = RL_CFG2;
1544 		sc->rl_cfg3 = RL_CFG3;
1545 		sc->rl_cfg4 = RL_CFG4;
1546 		sc->rl_cfg5 = RL_CFG5;
1547 	}
1548 
1549 	/* Reset the adapter. */
1550 	RL_LOCK(sc);
1551 	re_reset(sc);
1552 	RL_UNLOCK(sc);
1553 
1554 	/* Enable PME. */
1555 	CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
1556 	cfg = CSR_READ_1(sc, sc->rl_cfg1);
1557 	cfg |= RL_CFG1_PME;
1558 	CSR_WRITE_1(sc, sc->rl_cfg1, cfg);
1559 	cfg = CSR_READ_1(sc, sc->rl_cfg5);
1560 	cfg &= RL_CFG5_PME_STS;
1561 	CSR_WRITE_1(sc, sc->rl_cfg5, cfg);
1562 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
1563 
1564 	if ((sc->rl_flags & RL_FLAG_PAR) != 0) {
1565 		/*
1566 		 * XXX Should have a better way to extract station
1567 		 * address from EEPROM.
1568 		 */
1569 		for (i = 0; i < ETHER_ADDR_LEN; i++)
1570 			eaddr[i] = CSR_READ_1(sc, RL_IDR0 + i);
1571 	} else {
1572 		sc->rl_eewidth = RL_9356_ADDR_LEN;
1573 		re_read_eeprom(sc, (caddr_t)&re_did, 0, 1);
1574 		if (re_did != 0x8129)
1575 			sc->rl_eewidth = RL_9346_ADDR_LEN;
1576 
1577 		/*
1578 		 * Get station address from the EEPROM.
1579 		 */
1580 		re_read_eeprom(sc, (caddr_t)as, RL_EE_EADDR, 3);
1581 		for (i = 0; i < ETHER_ADDR_LEN / 2; i++)
1582 			as[i] = le16toh(as[i]);
1583 		bcopy(as, eaddr, ETHER_ADDR_LEN);
1584 	}
1585 
1586 	if (sc->rl_type == RL_8169) {
1587 		/* Set RX length mask and number of descriptors. */
1588 		sc->rl_rxlenmask = RL_RDESC_STAT_GFRAGLEN;
1589 		sc->rl_txstart = RL_GTXSTART;
1590 		sc->rl_ldata.rl_tx_desc_cnt = RL_8169_TX_DESC_CNT;
1591 		sc->rl_ldata.rl_rx_desc_cnt = RL_8169_RX_DESC_CNT;
1592 	} else {
1593 		/* Set RX length mask and number of descriptors. */
1594 		sc->rl_rxlenmask = RL_RDESC_STAT_FRAGLEN;
1595 		sc->rl_txstart = RL_TXSTART;
1596 		sc->rl_ldata.rl_tx_desc_cnt = RL_8139_TX_DESC_CNT;
1597 		sc->rl_ldata.rl_rx_desc_cnt = RL_8139_RX_DESC_CNT;
1598 	}
1599 
1600 	error = re_allocmem(dev, sc);
1601 	if (error)
1602 		goto fail;
1603 	re_add_sysctls(sc);
1604 
1605 	ifp = sc->rl_ifp = if_alloc(IFT_ETHER);
1606 	if (ifp == NULL) {
1607 		device_printf(dev, "can not if_alloc()\n");
1608 		error = ENOSPC;
1609 		goto fail;
1610 	}
1611 
1612 	/* Take controller out of deep sleep mode. */
1613 	if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) {
1614 		if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80)
1615 			CSR_WRITE_1(sc, RL_GPIO,
1616 			    CSR_READ_1(sc, RL_GPIO) | 0x01);
1617 		else
1618 			CSR_WRITE_1(sc, RL_GPIO,
1619 			    CSR_READ_1(sc, RL_GPIO) & ~0x01);
1620 	}
1621 
1622 	/* Take PHY out of power down mode. */
1623 	if ((sc->rl_flags & RL_FLAG_PHYWAKE_PM) != 0) {
1624 		CSR_WRITE_1(sc, RL_PMCH, CSR_READ_1(sc, RL_PMCH) | 0x80);
1625 		if (hw_rev->rl_rev == RL_HWREV_8401E)
1626 			CSR_WRITE_1(sc, 0xD1, CSR_READ_1(sc, 0xD1) & ~0x08);
1627 	}
1628 	if ((sc->rl_flags & RL_FLAG_PHYWAKE) != 0) {
1629 		re_gmii_writereg(dev, 1, 0x1f, 0);
1630 		re_gmii_writereg(dev, 1, 0x0e, 0);
1631 	}
1632 
1633 	ifp->if_softc = sc;
1634 	if_initname(ifp, device_get_name(dev), device_get_unit(dev));
1635 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
1636 	ifp->if_ioctl = re_ioctl;
1637 	ifp->if_start = re_start;
1638 	/*
1639 	 * RTL8168/8111C generates wrong IP checksummed frame if the
1640 	 * packet has IP options so disable TX checksum offloading.
1641 	 */
1642 	if (sc->rl_hwrev->rl_rev == RL_HWREV_8168C ||
1643 	    sc->rl_hwrev->rl_rev == RL_HWREV_8168C_SPIN2 ||
1644 	    sc->rl_hwrev->rl_rev == RL_HWREV_8168CP) {
1645 		ifp->if_hwassist = 0;
1646 		ifp->if_capabilities = IFCAP_RXCSUM | IFCAP_TSO4;
1647 	} else {
1648 		ifp->if_hwassist = CSUM_IP | CSUM_TCP | CSUM_UDP;
1649 		ifp->if_capabilities = IFCAP_HWCSUM | IFCAP_TSO4;
1650 	}
1651 	ifp->if_hwassist |= CSUM_TSO;
1652 	ifp->if_capenable = ifp->if_capabilities;
1653 	ifp->if_init = re_init;
1654 	IFQ_SET_MAXLEN(&ifp->if_snd, RL_IFQ_MAXLEN);
1655 	ifp->if_snd.ifq_drv_maxlen = RL_IFQ_MAXLEN;
1656 	IFQ_SET_READY(&ifp->if_snd);
1657 
1658 	TASK_INIT(&sc->rl_inttask, 0, re_int_task, sc);
1659 
1660 #define	RE_PHYAD_INTERNAL	 0
1661 
1662 	/* Do MII setup. */
1663 	phy = RE_PHYAD_INTERNAL;
1664 	if (sc->rl_type == RL_8169)
1665 		phy = 1;
1666 	capmask = BMSR_DEFCAPMASK;
1667 	if ((sc->rl_flags & RL_FLAG_FASTETHER) != 0)
1668 		 capmask &= ~BMSR_EXTSTAT;
1669 	error = mii_attach(dev, &sc->rl_miibus, ifp, re_ifmedia_upd,
1670 	    re_ifmedia_sts, capmask, phy, MII_OFFSET_ANY, MIIF_DOPAUSE);
1671 	if (error != 0) {
1672 		device_printf(dev, "attaching PHYs failed\n");
1673 		goto fail;
1674 	}
1675 
1676 	/*
1677 	 * Call MI attach routine.
1678 	 */
1679 	ether_ifattach(ifp, eaddr);
1680 
1681 	/* VLAN capability setup */
1682 	ifp->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
1683 	if (ifp->if_capabilities & IFCAP_HWCSUM)
1684 		ifp->if_capabilities |= IFCAP_VLAN_HWCSUM;
1685 	/* Enable WOL if PM is supported. */
1686 	if (pci_find_cap(sc->rl_dev, PCIY_PMG, &reg) == 0)
1687 		ifp->if_capabilities |= IFCAP_WOL;
1688 	ifp->if_capenable = ifp->if_capabilities;
1689 	ifp->if_capenable &= ~(IFCAP_WOL_UCAST | IFCAP_WOL_MCAST);
1690 	/*
1691 	 * Don't enable TSO by default.  It is known to generate
1692 	 * corrupted TCP segments(bad TCP options) under certain
1693 	 * circumstances.
1694 	 */
1695 	ifp->if_hwassist &= ~CSUM_TSO;
1696 	ifp->if_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
1697 #ifdef DEVICE_POLLING
1698 	ifp->if_capabilities |= IFCAP_POLLING;
1699 #endif
1700 	/*
1701 	 * Tell the upper layer(s) we support long frames.
1702 	 * Must appear after the call to ether_ifattach() because
1703 	 * ether_ifattach() sets ifi_hdrlen to the default value.
1704 	 */
1705 	ifp->if_hdrlen = sizeof(struct ether_vlan_header);
1706 
1707 #ifdef DEV_NETMAP
1708 	re_netmap_attach(sc);
1709 #endif /* DEV_NETMAP */
1710 
1711 #ifdef RE_DIAG
1712 	/*
1713 	 * Perform hardware diagnostic on the original RTL8169.
1714 	 * Some 32-bit cards were incorrectly wired and would
1715 	 * malfunction if plugged into a 64-bit slot.
1716 	 */
1717 	if (hwrev == RL_HWREV_8169) {
1718 		error = re_diag(sc);
1719 		if (error) {
1720 			device_printf(dev,
1721 		    	"attach aborted due to hardware diag failure\n");
1722 			ether_ifdetach(ifp);
1723 			goto fail;
1724 		}
1725 	}
1726 #endif
1727 
1728 #ifdef RE_TX_MODERATION
1729 	intr_filter = 1;
1730 #endif
1731 	/* Hook interrupt last to avoid having to lock softc */
1732 	if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) != 0 &&
1733 	    intr_filter == 0) {
1734 		error = bus_setup_intr(dev, sc->rl_irq[0],
1735 		    INTR_TYPE_NET | INTR_MPSAFE, NULL, re_intr_msi, sc,
1736 		    &sc->rl_intrhand[0]);
1737 	} else {
1738 		error = bus_setup_intr(dev, sc->rl_irq[0],
1739 		    INTR_TYPE_NET | INTR_MPSAFE, re_intr, NULL, sc,
1740 		    &sc->rl_intrhand[0]);
1741 	}
1742 	if (error) {
1743 		device_printf(dev, "couldn't set up irq\n");
1744 		ether_ifdetach(ifp);
1745 		goto fail;
1746 	}
1747 
1748 	NETDUMP_SET(ifp, re);
1749 
1750 fail:
1751 	if (error)
1752 		re_detach(dev);
1753 
1754 	return (error);
1755 }
1756 
1757 /*
1758  * Shutdown hardware and free up resources. This can be called any
1759  * time after the mutex has been initialized. It is called in both
1760  * the error case in attach and the normal detach case so it needs
1761  * to be careful about only freeing resources that have actually been
1762  * allocated.
1763  */
1764 static int
1765 re_detach(device_t dev)
1766 {
1767 	struct rl_softc		*sc;
1768 	struct ifnet		*ifp;
1769 	int			i, rid;
1770 
1771 	sc = device_get_softc(dev);
1772 	ifp = sc->rl_ifp;
1773 	KASSERT(mtx_initialized(&sc->rl_mtx), ("re mutex not initialized"));
1774 
1775 	/* These should only be active if attach succeeded */
1776 	if (device_is_attached(dev)) {
1777 #ifdef DEVICE_POLLING
1778 		if (ifp->if_capenable & IFCAP_POLLING)
1779 			ether_poll_deregister(ifp);
1780 #endif
1781 		RL_LOCK(sc);
1782 #if 0
1783 		sc->suspended = 1;
1784 #endif
1785 		re_stop(sc);
1786 		RL_UNLOCK(sc);
1787 		callout_drain(&sc->rl_stat_callout);
1788 		taskqueue_drain(taskqueue_fast, &sc->rl_inttask);
1789 		/*
1790 		 * Force off the IFF_UP flag here, in case someone
1791 		 * still had a BPF descriptor attached to this
1792 		 * interface. If they do, ether_ifdetach() will cause
1793 		 * the BPF code to try and clear the promisc mode
1794 		 * flag, which will bubble down to re_ioctl(),
1795 		 * which will try to call re_init() again. This will
1796 		 * turn the NIC back on and restart the MII ticker,
1797 		 * which will panic the system when the kernel tries
1798 		 * to invoke the re_tick() function that isn't there
1799 		 * anymore.
1800 		 */
1801 		ifp->if_flags &= ~IFF_UP;
1802 		ether_ifdetach(ifp);
1803 	}
1804 	if (sc->rl_miibus)
1805 		device_delete_child(dev, sc->rl_miibus);
1806 	bus_generic_detach(dev);
1807 
1808 	/*
1809 	 * The rest is resource deallocation, so we should already be
1810 	 * stopped here.
1811 	 */
1812 
1813 	if (sc->rl_intrhand[0] != NULL) {
1814 		bus_teardown_intr(dev, sc->rl_irq[0], sc->rl_intrhand[0]);
1815 		sc->rl_intrhand[0] = NULL;
1816 	}
1817 	if (ifp != NULL) {
1818 #ifdef DEV_NETMAP
1819 		netmap_detach(ifp);
1820 #endif /* DEV_NETMAP */
1821 		if_free(ifp);
1822 	}
1823 	if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) == 0)
1824 		rid = 0;
1825 	else
1826 		rid = 1;
1827 	if (sc->rl_irq[0] != NULL) {
1828 		bus_release_resource(dev, SYS_RES_IRQ, rid, sc->rl_irq[0]);
1829 		sc->rl_irq[0] = NULL;
1830 	}
1831 	if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) != 0)
1832 		pci_release_msi(dev);
1833 	if (sc->rl_res_pba) {
1834 		rid = PCIR_BAR(4);
1835 		bus_release_resource(dev, SYS_RES_MEMORY, rid, sc->rl_res_pba);
1836 	}
1837 	if (sc->rl_res)
1838 		bus_release_resource(dev, sc->rl_res_type, sc->rl_res_id,
1839 		    sc->rl_res);
1840 
1841 	/* Unload and free the RX DMA ring memory and map */
1842 
1843 	if (sc->rl_ldata.rl_rx_list_tag) {
1844 		if (sc->rl_ldata.rl_rx_list_addr)
1845 			bus_dmamap_unload(sc->rl_ldata.rl_rx_list_tag,
1846 			    sc->rl_ldata.rl_rx_list_map);
1847 		if (sc->rl_ldata.rl_rx_list)
1848 			bus_dmamem_free(sc->rl_ldata.rl_rx_list_tag,
1849 			    sc->rl_ldata.rl_rx_list,
1850 			    sc->rl_ldata.rl_rx_list_map);
1851 		bus_dma_tag_destroy(sc->rl_ldata.rl_rx_list_tag);
1852 	}
1853 
1854 	/* Unload and free the TX DMA ring memory and map */
1855 
1856 	if (sc->rl_ldata.rl_tx_list_tag) {
1857 		if (sc->rl_ldata.rl_tx_list_addr)
1858 			bus_dmamap_unload(sc->rl_ldata.rl_tx_list_tag,
1859 			    sc->rl_ldata.rl_tx_list_map);
1860 		if (sc->rl_ldata.rl_tx_list)
1861 			bus_dmamem_free(sc->rl_ldata.rl_tx_list_tag,
1862 			    sc->rl_ldata.rl_tx_list,
1863 			    sc->rl_ldata.rl_tx_list_map);
1864 		bus_dma_tag_destroy(sc->rl_ldata.rl_tx_list_tag);
1865 	}
1866 
1867 	/* Destroy all the RX and TX buffer maps */
1868 
1869 	if (sc->rl_ldata.rl_tx_mtag) {
1870 		for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) {
1871 			if (sc->rl_ldata.rl_tx_desc[i].tx_dmamap)
1872 				bus_dmamap_destroy(sc->rl_ldata.rl_tx_mtag,
1873 				    sc->rl_ldata.rl_tx_desc[i].tx_dmamap);
1874 		}
1875 		bus_dma_tag_destroy(sc->rl_ldata.rl_tx_mtag);
1876 	}
1877 	if (sc->rl_ldata.rl_rx_mtag) {
1878 		for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
1879 			if (sc->rl_ldata.rl_rx_desc[i].rx_dmamap)
1880 				bus_dmamap_destroy(sc->rl_ldata.rl_rx_mtag,
1881 				    sc->rl_ldata.rl_rx_desc[i].rx_dmamap);
1882 		}
1883 		if (sc->rl_ldata.rl_rx_sparemap)
1884 			bus_dmamap_destroy(sc->rl_ldata.rl_rx_mtag,
1885 			    sc->rl_ldata.rl_rx_sparemap);
1886 		bus_dma_tag_destroy(sc->rl_ldata.rl_rx_mtag);
1887 	}
1888 	if (sc->rl_ldata.rl_jrx_mtag) {
1889 		for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
1890 			if (sc->rl_ldata.rl_jrx_desc[i].rx_dmamap)
1891 				bus_dmamap_destroy(sc->rl_ldata.rl_jrx_mtag,
1892 				    sc->rl_ldata.rl_jrx_desc[i].rx_dmamap);
1893 		}
1894 		if (sc->rl_ldata.rl_jrx_sparemap)
1895 			bus_dmamap_destroy(sc->rl_ldata.rl_jrx_mtag,
1896 			    sc->rl_ldata.rl_jrx_sparemap);
1897 		bus_dma_tag_destroy(sc->rl_ldata.rl_jrx_mtag);
1898 	}
1899 	/* Unload and free the stats buffer and map */
1900 
1901 	if (sc->rl_ldata.rl_stag) {
1902 		if (sc->rl_ldata.rl_stats_addr)
1903 			bus_dmamap_unload(sc->rl_ldata.rl_stag,
1904 			    sc->rl_ldata.rl_smap);
1905 		if (sc->rl_ldata.rl_stats)
1906 			bus_dmamem_free(sc->rl_ldata.rl_stag,
1907 			    sc->rl_ldata.rl_stats, sc->rl_ldata.rl_smap);
1908 		bus_dma_tag_destroy(sc->rl_ldata.rl_stag);
1909 	}
1910 
1911 	if (sc->rl_parent_tag)
1912 		bus_dma_tag_destroy(sc->rl_parent_tag);
1913 
1914 	mtx_destroy(&sc->rl_mtx);
1915 
1916 	return (0);
1917 }
1918 
1919 static __inline void
1920 re_discard_rxbuf(struct rl_softc *sc, int idx)
1921 {
1922 	struct rl_desc		*desc;
1923 	struct rl_rxdesc	*rxd;
1924 	uint32_t		cmdstat;
1925 
1926 	if (sc->rl_ifp->if_mtu > RL_MTU &&
1927 	    (sc->rl_flags & RL_FLAG_JUMBOV2) != 0)
1928 		rxd = &sc->rl_ldata.rl_jrx_desc[idx];
1929 	else
1930 		rxd = &sc->rl_ldata.rl_rx_desc[idx];
1931 	desc = &sc->rl_ldata.rl_rx_list[idx];
1932 	desc->rl_vlanctl = 0;
1933 	cmdstat = rxd->rx_size;
1934 	if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1)
1935 		cmdstat |= RL_RDESC_CMD_EOR;
1936 	desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN);
1937 }
1938 
1939 static int
1940 re_newbuf(struct rl_softc *sc, int idx)
1941 {
1942 	struct mbuf		*m;
1943 	struct rl_rxdesc	*rxd;
1944 	bus_dma_segment_t	segs[1];
1945 	bus_dmamap_t		map;
1946 	struct rl_desc		*desc;
1947 	uint32_t		cmdstat;
1948 	int			error, nsegs;
1949 
1950 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1951 	if (m == NULL)
1952 		return (ENOBUFS);
1953 
1954 	m->m_len = m->m_pkthdr.len = MCLBYTES;
1955 #ifdef RE_FIXUP_RX
1956 	/*
1957 	 * This is part of an evil trick to deal with non-x86 platforms.
1958 	 * The RealTek chip requires RX buffers to be aligned on 64-bit
1959 	 * boundaries, but that will hose non-x86 machines. To get around
1960 	 * this, we leave some empty space at the start of each buffer
1961 	 * and for non-x86 hosts, we copy the buffer back six bytes
1962 	 * to achieve word alignment. This is slightly more efficient
1963 	 * than allocating a new buffer, copying the contents, and
1964 	 * discarding the old buffer.
1965 	 */
1966 	m_adj(m, RE_ETHER_ALIGN);
1967 #endif
1968 	error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_rx_mtag,
1969 	    sc->rl_ldata.rl_rx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT);
1970 	if (error != 0) {
1971 		m_freem(m);
1972 		return (ENOBUFS);
1973 	}
1974 	KASSERT(nsegs == 1, ("%s: %d segment returned!", __func__, nsegs));
1975 
1976 	rxd = &sc->rl_ldata.rl_rx_desc[idx];
1977 	if (rxd->rx_m != NULL) {
1978 		bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap,
1979 		    BUS_DMASYNC_POSTREAD);
1980 		bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap);
1981 	}
1982 
1983 	rxd->rx_m = m;
1984 	map = rxd->rx_dmamap;
1985 	rxd->rx_dmamap = sc->rl_ldata.rl_rx_sparemap;
1986 	rxd->rx_size = segs[0].ds_len;
1987 	sc->rl_ldata.rl_rx_sparemap = map;
1988 	bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag, rxd->rx_dmamap,
1989 	    BUS_DMASYNC_PREREAD);
1990 
1991 	desc = &sc->rl_ldata.rl_rx_list[idx];
1992 	desc->rl_vlanctl = 0;
1993 	desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[0].ds_addr));
1994 	desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[0].ds_addr));
1995 	cmdstat = segs[0].ds_len;
1996 	if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1)
1997 		cmdstat |= RL_RDESC_CMD_EOR;
1998 	desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN);
1999 
2000 	return (0);
2001 }
2002 
2003 static int
2004 re_jumbo_newbuf(struct rl_softc *sc, int idx)
2005 {
2006 	struct mbuf		*m;
2007 	struct rl_rxdesc	*rxd;
2008 	bus_dma_segment_t	segs[1];
2009 	bus_dmamap_t		map;
2010 	struct rl_desc		*desc;
2011 	uint32_t		cmdstat;
2012 	int			error, nsegs;
2013 
2014 	m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUM9BYTES);
2015 	if (m == NULL)
2016 		return (ENOBUFS);
2017 	m->m_len = m->m_pkthdr.len = MJUM9BYTES;
2018 #ifdef RE_FIXUP_RX
2019 	m_adj(m, RE_ETHER_ALIGN);
2020 #endif
2021 	error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_jrx_mtag,
2022 	    sc->rl_ldata.rl_jrx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT);
2023 	if (error != 0) {
2024 		m_freem(m);
2025 		return (ENOBUFS);
2026 	}
2027 	KASSERT(nsegs == 1, ("%s: %d segment returned!", __func__, nsegs));
2028 
2029 	rxd = &sc->rl_ldata.rl_jrx_desc[idx];
2030 	if (rxd->rx_m != NULL) {
2031 		bus_dmamap_sync(sc->rl_ldata.rl_jrx_mtag, rxd->rx_dmamap,
2032 		    BUS_DMASYNC_POSTREAD);
2033 		bus_dmamap_unload(sc->rl_ldata.rl_jrx_mtag, rxd->rx_dmamap);
2034 	}
2035 
2036 	rxd->rx_m = m;
2037 	map = rxd->rx_dmamap;
2038 	rxd->rx_dmamap = sc->rl_ldata.rl_jrx_sparemap;
2039 	rxd->rx_size = segs[0].ds_len;
2040 	sc->rl_ldata.rl_jrx_sparemap = map;
2041 	bus_dmamap_sync(sc->rl_ldata.rl_jrx_mtag, rxd->rx_dmamap,
2042 	    BUS_DMASYNC_PREREAD);
2043 
2044 	desc = &sc->rl_ldata.rl_rx_list[idx];
2045 	desc->rl_vlanctl = 0;
2046 	desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[0].ds_addr));
2047 	desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[0].ds_addr));
2048 	cmdstat = segs[0].ds_len;
2049 	if (idx == sc->rl_ldata.rl_rx_desc_cnt - 1)
2050 		cmdstat |= RL_RDESC_CMD_EOR;
2051 	desc->rl_cmdstat = htole32(cmdstat | RL_RDESC_CMD_OWN);
2052 
2053 	return (0);
2054 }
2055 
2056 #ifdef RE_FIXUP_RX
2057 static __inline void
2058 re_fixup_rx(struct mbuf *m)
2059 {
2060 	int                     i;
2061 	uint16_t                *src, *dst;
2062 
2063 	src = mtod(m, uint16_t *);
2064 	dst = src - (RE_ETHER_ALIGN - ETHER_ALIGN) / sizeof *src;
2065 
2066 	for (i = 0; i < (m->m_len / sizeof(uint16_t) + 1); i++)
2067 		*dst++ = *src++;
2068 
2069 	m->m_data -= RE_ETHER_ALIGN - ETHER_ALIGN;
2070 }
2071 #endif
2072 
2073 static int
2074 re_tx_list_init(struct rl_softc *sc)
2075 {
2076 	struct rl_desc		*desc;
2077 	int			i;
2078 
2079 	RL_LOCK_ASSERT(sc);
2080 
2081 	bzero(sc->rl_ldata.rl_tx_list,
2082 	    sc->rl_ldata.rl_tx_desc_cnt * sizeof(struct rl_desc));
2083 	for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++)
2084 		sc->rl_ldata.rl_tx_desc[i].tx_m = NULL;
2085 #ifdef DEV_NETMAP
2086 	re_netmap_tx_init(sc);
2087 #endif /* DEV_NETMAP */
2088 	/* Set EOR. */
2089 	desc = &sc->rl_ldata.rl_tx_list[sc->rl_ldata.rl_tx_desc_cnt - 1];
2090 	desc->rl_cmdstat |= htole32(RL_TDESC_CMD_EOR);
2091 
2092 	bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag,
2093 	    sc->rl_ldata.rl_tx_list_map,
2094 	    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
2095 
2096 	sc->rl_ldata.rl_tx_prodidx = 0;
2097 	sc->rl_ldata.rl_tx_considx = 0;
2098 	sc->rl_ldata.rl_tx_free = sc->rl_ldata.rl_tx_desc_cnt;
2099 
2100 	return (0);
2101 }
2102 
2103 static int
2104 re_rx_list_init(struct rl_softc *sc)
2105 {
2106 	int			error, i;
2107 
2108 	bzero(sc->rl_ldata.rl_rx_list,
2109 	    sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc));
2110 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
2111 		sc->rl_ldata.rl_rx_desc[i].rx_m = NULL;
2112 		if ((error = re_newbuf(sc, i)) != 0)
2113 			return (error);
2114 	}
2115 #ifdef DEV_NETMAP
2116 	re_netmap_rx_init(sc);
2117 #endif /* DEV_NETMAP */
2118 
2119 	/* Flush the RX descriptors */
2120 
2121 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
2122 	    sc->rl_ldata.rl_rx_list_map,
2123 	    BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD);
2124 
2125 	sc->rl_ldata.rl_rx_prodidx = 0;
2126 	sc->rl_head = sc->rl_tail = NULL;
2127 	sc->rl_int_rx_act = 0;
2128 
2129 	return (0);
2130 }
2131 
2132 static int
2133 re_jrx_list_init(struct rl_softc *sc)
2134 {
2135 	int			error, i;
2136 
2137 	bzero(sc->rl_ldata.rl_rx_list,
2138 	    sc->rl_ldata.rl_rx_desc_cnt * sizeof(struct rl_desc));
2139 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
2140 		sc->rl_ldata.rl_jrx_desc[i].rx_m = NULL;
2141 		if ((error = re_jumbo_newbuf(sc, i)) != 0)
2142 			return (error);
2143 	}
2144 
2145 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
2146 	    sc->rl_ldata.rl_rx_list_map,
2147 	    BUS_DMASYNC_PREWRITE | BUS_DMASYNC_PREREAD);
2148 
2149 	sc->rl_ldata.rl_rx_prodidx = 0;
2150 	sc->rl_head = sc->rl_tail = NULL;
2151 	sc->rl_int_rx_act = 0;
2152 
2153 	return (0);
2154 }
2155 
2156 /*
2157  * RX handler for C+ and 8169. For the gigE chips, we support
2158  * the reception of jumbo frames that have been fragmented
2159  * across multiple 2K mbuf cluster buffers.
2160  */
2161 static int
2162 re_rxeof(struct rl_softc *sc, int *rx_npktsp)
2163 {
2164 	struct mbuf		*m;
2165 	struct ifnet		*ifp;
2166 	int			i, rxerr, total_len;
2167 	struct rl_desc		*cur_rx;
2168 	u_int32_t		rxstat, rxvlan;
2169 	int			jumbo, maxpkt = 16, rx_npkts = 0;
2170 
2171 	RL_LOCK_ASSERT(sc);
2172 
2173 	ifp = sc->rl_ifp;
2174 #ifdef DEV_NETMAP
2175 	if (netmap_rx_irq(ifp, 0, &rx_npkts))
2176 		return 0;
2177 #endif /* DEV_NETMAP */
2178 	if (ifp->if_mtu > RL_MTU && (sc->rl_flags & RL_FLAG_JUMBOV2) != 0)
2179 		jumbo = 1;
2180 	else
2181 		jumbo = 0;
2182 
2183 	/* Invalidate the descriptor memory */
2184 
2185 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
2186 	    sc->rl_ldata.rl_rx_list_map,
2187 	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
2188 
2189 	for (i = sc->rl_ldata.rl_rx_prodidx; maxpkt > 0;
2190 	    i = RL_RX_DESC_NXT(sc, i)) {
2191 		if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2192 			break;
2193 		cur_rx = &sc->rl_ldata.rl_rx_list[i];
2194 		rxstat = le32toh(cur_rx->rl_cmdstat);
2195 		if ((rxstat & RL_RDESC_STAT_OWN) != 0)
2196 			break;
2197 		total_len = rxstat & sc->rl_rxlenmask;
2198 		rxvlan = le32toh(cur_rx->rl_vlanctl);
2199 		if (jumbo != 0)
2200 			m = sc->rl_ldata.rl_jrx_desc[i].rx_m;
2201 		else
2202 			m = sc->rl_ldata.rl_rx_desc[i].rx_m;
2203 
2204 		if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0 &&
2205 		    (rxstat & (RL_RDESC_STAT_SOF | RL_RDESC_STAT_EOF)) !=
2206 		    (RL_RDESC_STAT_SOF | RL_RDESC_STAT_EOF)) {
2207 			/*
2208 			 * RTL8168C or later controllers do not
2209 			 * support multi-fragment packet.
2210 			 */
2211 			re_discard_rxbuf(sc, i);
2212 			continue;
2213 		} else if ((rxstat & RL_RDESC_STAT_EOF) == 0) {
2214 			if (re_newbuf(sc, i) != 0) {
2215 				/*
2216 				 * If this is part of a multi-fragment packet,
2217 				 * discard all the pieces.
2218 				 */
2219 				if (sc->rl_head != NULL) {
2220 					m_freem(sc->rl_head);
2221 					sc->rl_head = sc->rl_tail = NULL;
2222 				}
2223 				re_discard_rxbuf(sc, i);
2224 				continue;
2225 			}
2226 			m->m_len = RE_RX_DESC_BUFLEN;
2227 			if (sc->rl_head == NULL)
2228 				sc->rl_head = sc->rl_tail = m;
2229 			else {
2230 				m->m_flags &= ~M_PKTHDR;
2231 				sc->rl_tail->m_next = m;
2232 				sc->rl_tail = m;
2233 			}
2234 			continue;
2235 		}
2236 
2237 		/*
2238 		 * NOTE: for the 8139C+, the frame length field
2239 		 * is always 12 bits in size, but for the gigE chips,
2240 		 * it is 13 bits (since the max RX frame length is 16K).
2241 		 * Unfortunately, all 32 bits in the status word
2242 		 * were already used, so to make room for the extra
2243 		 * length bit, RealTek took out the 'frame alignment
2244 		 * error' bit and shifted the other status bits
2245 		 * over one slot. The OWN, EOR, FS and LS bits are
2246 		 * still in the same places. We have already extracted
2247 		 * the frame length and checked the OWN bit, so rather
2248 		 * than using an alternate bit mapping, we shift the
2249 		 * status bits one space to the right so we can evaluate
2250 		 * them using the 8169 status as though it was in the
2251 		 * same format as that of the 8139C+.
2252 		 */
2253 		if (sc->rl_type == RL_8169)
2254 			rxstat >>= 1;
2255 
2256 		/*
2257 		 * if total_len > 2^13-1, both _RXERRSUM and _GIANT will be
2258 		 * set, but if CRC is clear, it will still be a valid frame.
2259 		 */
2260 		if ((rxstat & RL_RDESC_STAT_RXERRSUM) != 0) {
2261 			rxerr = 1;
2262 			if ((sc->rl_flags & RL_FLAG_JUMBOV2) == 0 &&
2263 			    total_len > 8191 &&
2264 			    (rxstat & RL_RDESC_STAT_ERRS) == RL_RDESC_STAT_GIANT)
2265 				rxerr = 0;
2266 			if (rxerr != 0) {
2267 				if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
2268 				/*
2269 				 * If this is part of a multi-fragment packet,
2270 				 * discard all the pieces.
2271 				 */
2272 				if (sc->rl_head != NULL) {
2273 					m_freem(sc->rl_head);
2274 					sc->rl_head = sc->rl_tail = NULL;
2275 				}
2276 				re_discard_rxbuf(sc, i);
2277 				continue;
2278 			}
2279 		}
2280 
2281 		/*
2282 		 * If allocating a replacement mbuf fails,
2283 		 * reload the current one.
2284 		 */
2285 		if (jumbo != 0)
2286 			rxerr = re_jumbo_newbuf(sc, i);
2287 		else
2288 			rxerr = re_newbuf(sc, i);
2289 		if (rxerr != 0) {
2290 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
2291 			if (sc->rl_head != NULL) {
2292 				m_freem(sc->rl_head);
2293 				sc->rl_head = sc->rl_tail = NULL;
2294 			}
2295 			re_discard_rxbuf(sc, i);
2296 			continue;
2297 		}
2298 
2299 		if (sc->rl_head != NULL) {
2300 			if (jumbo != 0)
2301 				m->m_len = total_len;
2302 			else {
2303 				m->m_len = total_len % RE_RX_DESC_BUFLEN;
2304 				if (m->m_len == 0)
2305 					m->m_len = RE_RX_DESC_BUFLEN;
2306 			}
2307 			/*
2308 			 * Special case: if there's 4 bytes or less
2309 			 * in this buffer, the mbuf can be discarded:
2310 			 * the last 4 bytes is the CRC, which we don't
2311 			 * care about anyway.
2312 			 */
2313 			if (m->m_len <= ETHER_CRC_LEN) {
2314 				sc->rl_tail->m_len -=
2315 				    (ETHER_CRC_LEN - m->m_len);
2316 				m_freem(m);
2317 			} else {
2318 				m->m_len -= ETHER_CRC_LEN;
2319 				m->m_flags &= ~M_PKTHDR;
2320 				sc->rl_tail->m_next = m;
2321 			}
2322 			m = sc->rl_head;
2323 			sc->rl_head = sc->rl_tail = NULL;
2324 			m->m_pkthdr.len = total_len - ETHER_CRC_LEN;
2325 		} else
2326 			m->m_pkthdr.len = m->m_len =
2327 			    (total_len - ETHER_CRC_LEN);
2328 
2329 #ifdef RE_FIXUP_RX
2330 		re_fixup_rx(m);
2331 #endif
2332 		if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2333 		m->m_pkthdr.rcvif = ifp;
2334 
2335 		/* Do RX checksumming if enabled */
2336 
2337 		if (ifp->if_capenable & IFCAP_RXCSUM) {
2338 			if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) {
2339 				/* Check IP header checksum */
2340 				if (rxstat & RL_RDESC_STAT_PROTOID)
2341 					m->m_pkthdr.csum_flags |=
2342 					    CSUM_IP_CHECKED;
2343 				if (!(rxstat & RL_RDESC_STAT_IPSUMBAD))
2344 					m->m_pkthdr.csum_flags |=
2345 					    CSUM_IP_VALID;
2346 
2347 				/* Check TCP/UDP checksum */
2348 				if ((RL_TCPPKT(rxstat) &&
2349 				    !(rxstat & RL_RDESC_STAT_TCPSUMBAD)) ||
2350 				    (RL_UDPPKT(rxstat) &&
2351 				     !(rxstat & RL_RDESC_STAT_UDPSUMBAD))) {
2352 					m->m_pkthdr.csum_flags |=
2353 						CSUM_DATA_VALID|CSUM_PSEUDO_HDR;
2354 					m->m_pkthdr.csum_data = 0xffff;
2355 				}
2356 			} else {
2357 				/*
2358 				 * RTL8168C/RTL816CP/RTL8111C/RTL8111CP
2359 				 */
2360 				if ((rxstat & RL_RDESC_STAT_PROTOID) &&
2361 				    (rxvlan & RL_RDESC_IPV4))
2362 					m->m_pkthdr.csum_flags |=
2363 					    CSUM_IP_CHECKED;
2364 				if (!(rxstat & RL_RDESC_STAT_IPSUMBAD) &&
2365 				    (rxvlan & RL_RDESC_IPV4))
2366 					m->m_pkthdr.csum_flags |=
2367 					    CSUM_IP_VALID;
2368 				if (((rxstat & RL_RDESC_STAT_TCP) &&
2369 				    !(rxstat & RL_RDESC_STAT_TCPSUMBAD)) ||
2370 				    ((rxstat & RL_RDESC_STAT_UDP) &&
2371 				    !(rxstat & RL_RDESC_STAT_UDPSUMBAD))) {
2372 					m->m_pkthdr.csum_flags |=
2373 						CSUM_DATA_VALID|CSUM_PSEUDO_HDR;
2374 					m->m_pkthdr.csum_data = 0xffff;
2375 				}
2376 			}
2377 		}
2378 		maxpkt--;
2379 		if (rxvlan & RL_RDESC_VLANCTL_TAG) {
2380 			m->m_pkthdr.ether_vtag =
2381 			    bswap16((rxvlan & RL_RDESC_VLANCTL_DATA));
2382 			m->m_flags |= M_VLANTAG;
2383 		}
2384 		RL_UNLOCK(sc);
2385 		(*ifp->if_input)(ifp, m);
2386 		RL_LOCK(sc);
2387 		rx_npkts++;
2388 	}
2389 
2390 	/* Flush the RX DMA ring */
2391 
2392 	bus_dmamap_sync(sc->rl_ldata.rl_rx_list_tag,
2393 	    sc->rl_ldata.rl_rx_list_map,
2394 	    BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD);
2395 
2396 	sc->rl_ldata.rl_rx_prodidx = i;
2397 
2398 	if (rx_npktsp != NULL)
2399 		*rx_npktsp = rx_npkts;
2400 	if (maxpkt)
2401 		return (EAGAIN);
2402 
2403 	return (0);
2404 }
2405 
2406 static void
2407 re_txeof(struct rl_softc *sc)
2408 {
2409 	struct ifnet		*ifp;
2410 	struct rl_txdesc	*txd;
2411 	u_int32_t		txstat;
2412 	int			cons;
2413 
2414 	cons = sc->rl_ldata.rl_tx_considx;
2415 	if (cons == sc->rl_ldata.rl_tx_prodidx)
2416 		return;
2417 
2418 	ifp = sc->rl_ifp;
2419 #ifdef DEV_NETMAP
2420 	if (netmap_tx_irq(ifp, 0))
2421 		return;
2422 #endif /* DEV_NETMAP */
2423 	/* Invalidate the TX descriptor list */
2424 	bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag,
2425 	    sc->rl_ldata.rl_tx_list_map,
2426 	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
2427 
2428 	for (; cons != sc->rl_ldata.rl_tx_prodidx;
2429 	    cons = RL_TX_DESC_NXT(sc, cons)) {
2430 		txstat = le32toh(sc->rl_ldata.rl_tx_list[cons].rl_cmdstat);
2431 		if (txstat & RL_TDESC_STAT_OWN)
2432 			break;
2433 		/*
2434 		 * We only stash mbufs in the last descriptor
2435 		 * in a fragment chain, which also happens to
2436 		 * be the only place where the TX status bits
2437 		 * are valid.
2438 		 */
2439 		if (txstat & RL_TDESC_CMD_EOF) {
2440 			txd = &sc->rl_ldata.rl_tx_desc[cons];
2441 			bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag,
2442 			    txd->tx_dmamap, BUS_DMASYNC_POSTWRITE);
2443 			bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag,
2444 			    txd->tx_dmamap);
2445 			KASSERT(txd->tx_m != NULL,
2446 			    ("%s: freeing NULL mbufs!", __func__));
2447 			m_freem(txd->tx_m);
2448 			txd->tx_m = NULL;
2449 			if (txstat & (RL_TDESC_STAT_EXCESSCOL|
2450 			    RL_TDESC_STAT_COLCNT))
2451 				if_inc_counter(ifp, IFCOUNTER_COLLISIONS, 1);
2452 			if (txstat & RL_TDESC_STAT_TXERRSUM)
2453 				if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2454 			else
2455 				if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2456 		}
2457 		sc->rl_ldata.rl_tx_free++;
2458 		ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2459 	}
2460 	sc->rl_ldata.rl_tx_considx = cons;
2461 
2462 	/* No changes made to the TX ring, so no flush needed */
2463 
2464 	if (sc->rl_ldata.rl_tx_free != sc->rl_ldata.rl_tx_desc_cnt) {
2465 #ifdef RE_TX_MODERATION
2466 		/*
2467 		 * If not all descriptors have been reaped yet, reload
2468 		 * the timer so that we will eventually get another
2469 		 * interrupt that will cause us to re-enter this routine.
2470 		 * This is done in case the transmitter has gone idle.
2471 		 */
2472 		CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2473 #endif
2474 	} else
2475 		sc->rl_watchdog_timer = 0;
2476 }
2477 
2478 static void
2479 re_tick(void *xsc)
2480 {
2481 	struct rl_softc		*sc;
2482 	struct mii_data		*mii;
2483 
2484 	sc = xsc;
2485 
2486 	RL_LOCK_ASSERT(sc);
2487 
2488 	mii = device_get_softc(sc->rl_miibus);
2489 	mii_tick(mii);
2490 	if ((sc->rl_flags & RL_FLAG_LINK) == 0)
2491 		re_miibus_statchg(sc->rl_dev);
2492 	/*
2493 	 * Reclaim transmitted frames here. Technically it is not
2494 	 * necessary to do here but it ensures periodic reclamation
2495 	 * regardless of Tx completion interrupt which seems to be
2496 	 * lost on PCIe based controllers under certain situations.
2497 	 */
2498 	re_txeof(sc);
2499 	re_watchdog(sc);
2500 	callout_reset(&sc->rl_stat_callout, hz, re_tick, sc);
2501 }
2502 
2503 #ifdef DEVICE_POLLING
2504 static int
2505 re_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
2506 {
2507 	struct rl_softc *sc = ifp->if_softc;
2508 	int rx_npkts = 0;
2509 
2510 	RL_LOCK(sc);
2511 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2512 		rx_npkts = re_poll_locked(ifp, cmd, count);
2513 	RL_UNLOCK(sc);
2514 	return (rx_npkts);
2515 }
2516 
2517 static int
2518 re_poll_locked(struct ifnet *ifp, enum poll_cmd cmd, int count)
2519 {
2520 	struct rl_softc *sc = ifp->if_softc;
2521 	int rx_npkts;
2522 
2523 	RL_LOCK_ASSERT(sc);
2524 
2525 	sc->rxcycles = count;
2526 	re_rxeof(sc, &rx_npkts);
2527 	re_txeof(sc);
2528 
2529 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2530 		re_start_locked(ifp);
2531 
2532 	if (cmd == POLL_AND_CHECK_STATUS) { /* also check status register */
2533 		u_int16_t       status;
2534 
2535 		status = CSR_READ_2(sc, RL_ISR);
2536 		if (status == 0xffff)
2537 			return (rx_npkts);
2538 		if (status)
2539 			CSR_WRITE_2(sc, RL_ISR, status);
2540 		if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) &&
2541 		    (sc->rl_flags & RL_FLAG_PCIE))
2542 			CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2543 
2544 		/*
2545 		 * XXX check behaviour on receiver stalls.
2546 		 */
2547 
2548 		if (status & RL_ISR_SYSTEM_ERR) {
2549 			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2550 			re_init_locked(sc);
2551 		}
2552 	}
2553 	return (rx_npkts);
2554 }
2555 #endif /* DEVICE_POLLING */
2556 
2557 static int
2558 re_intr(void *arg)
2559 {
2560 	struct rl_softc		*sc;
2561 	uint16_t		status;
2562 
2563 	sc = arg;
2564 
2565 	status = CSR_READ_2(sc, RL_ISR);
2566 	if (status == 0xFFFF || (status & RL_INTRS_CPLUS) == 0)
2567                 return (FILTER_STRAY);
2568 	CSR_WRITE_2(sc, RL_IMR, 0);
2569 
2570 	taskqueue_enqueue(taskqueue_fast, &sc->rl_inttask);
2571 
2572 	return (FILTER_HANDLED);
2573 }
2574 
2575 static void
2576 re_int_task(void *arg, int npending)
2577 {
2578 	struct rl_softc		*sc;
2579 	struct ifnet		*ifp;
2580 	u_int16_t		status;
2581 	int			rval = 0;
2582 
2583 	sc = arg;
2584 	ifp = sc->rl_ifp;
2585 
2586 	RL_LOCK(sc);
2587 
2588 	status = CSR_READ_2(sc, RL_ISR);
2589         CSR_WRITE_2(sc, RL_ISR, status);
2590 
2591 	if (sc->suspended ||
2592 	    (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2593 		RL_UNLOCK(sc);
2594 		return;
2595 	}
2596 
2597 #ifdef DEVICE_POLLING
2598 	if  (ifp->if_capenable & IFCAP_POLLING) {
2599 		RL_UNLOCK(sc);
2600 		return;
2601 	}
2602 #endif
2603 
2604 	if (status & (RL_ISR_RX_OK|RL_ISR_RX_ERR|RL_ISR_FIFO_OFLOW))
2605 		rval = re_rxeof(sc, NULL);
2606 
2607 	/*
2608 	 * Some chips will ignore a second TX request issued
2609 	 * while an existing transmission is in progress. If
2610 	 * the transmitter goes idle but there are still
2611 	 * packets waiting to be sent, we need to restart the
2612 	 * channel here to flush them out. This only seems to
2613 	 * be required with the PCIe devices.
2614 	 */
2615 	if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) &&
2616 	    (sc->rl_flags & RL_FLAG_PCIE))
2617 		CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2618 	if (status & (
2619 #ifdef RE_TX_MODERATION
2620 	    RL_ISR_TIMEOUT_EXPIRED|
2621 #else
2622 	    RL_ISR_TX_OK|
2623 #endif
2624 	    RL_ISR_TX_ERR|RL_ISR_TX_DESC_UNAVAIL))
2625 		re_txeof(sc);
2626 
2627 	if (status & RL_ISR_SYSTEM_ERR) {
2628 		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2629 		re_init_locked(sc);
2630 	}
2631 
2632 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2633 		re_start_locked(ifp);
2634 
2635 	RL_UNLOCK(sc);
2636 
2637         if ((CSR_READ_2(sc, RL_ISR) & RL_INTRS_CPLUS) || rval) {
2638 		taskqueue_enqueue(taskqueue_fast, &sc->rl_inttask);
2639 		return;
2640 	}
2641 
2642 	CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS);
2643 }
2644 
2645 static void
2646 re_intr_msi(void *xsc)
2647 {
2648 	struct rl_softc		*sc;
2649 	struct ifnet		*ifp;
2650 	uint16_t		intrs, status;
2651 
2652 	sc = xsc;
2653 	RL_LOCK(sc);
2654 
2655 	ifp = sc->rl_ifp;
2656 #ifdef DEVICE_POLLING
2657 	if (ifp->if_capenable & IFCAP_POLLING) {
2658 		RL_UNLOCK(sc);
2659 		return;
2660 	}
2661 #endif
2662 	/* Disable interrupts. */
2663 	CSR_WRITE_2(sc, RL_IMR, 0);
2664 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2665 		RL_UNLOCK(sc);
2666 		return;
2667 	}
2668 
2669 	intrs = RL_INTRS_CPLUS;
2670 	status = CSR_READ_2(sc, RL_ISR);
2671         CSR_WRITE_2(sc, RL_ISR, status);
2672 	if (sc->rl_int_rx_act > 0) {
2673 		intrs &= ~(RL_ISR_RX_OK | RL_ISR_RX_ERR | RL_ISR_FIFO_OFLOW |
2674 		    RL_ISR_RX_OVERRUN);
2675 		status &= ~(RL_ISR_RX_OK | RL_ISR_RX_ERR | RL_ISR_FIFO_OFLOW |
2676 		    RL_ISR_RX_OVERRUN);
2677 	}
2678 
2679 	if (status & (RL_ISR_TIMEOUT_EXPIRED | RL_ISR_RX_OK | RL_ISR_RX_ERR |
2680 	    RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN)) {
2681 		re_rxeof(sc, NULL);
2682 		if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
2683 			if (sc->rl_int_rx_mod != 0 &&
2684 			    (status & (RL_ISR_RX_OK | RL_ISR_RX_ERR |
2685 			    RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN)) != 0) {
2686 				/* Rearm one-shot timer. */
2687 				CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2688 				intrs &= ~(RL_ISR_RX_OK | RL_ISR_RX_ERR |
2689 				    RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN);
2690 				sc->rl_int_rx_act = 1;
2691 			} else {
2692 				intrs |= RL_ISR_RX_OK | RL_ISR_RX_ERR |
2693 				    RL_ISR_FIFO_OFLOW | RL_ISR_RX_OVERRUN;
2694 				sc->rl_int_rx_act = 0;
2695 			}
2696 		}
2697 	}
2698 
2699 	/*
2700 	 * Some chips will ignore a second TX request issued
2701 	 * while an existing transmission is in progress. If
2702 	 * the transmitter goes idle but there are still
2703 	 * packets waiting to be sent, we need to restart the
2704 	 * channel here to flush them out. This only seems to
2705 	 * be required with the PCIe devices.
2706 	 */
2707 	if ((status & (RL_ISR_TX_OK | RL_ISR_TX_DESC_UNAVAIL)) &&
2708 	    (sc->rl_flags & RL_FLAG_PCIE))
2709 		CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2710 	if (status & (RL_ISR_TX_OK | RL_ISR_TX_ERR | RL_ISR_TX_DESC_UNAVAIL))
2711 		re_txeof(sc);
2712 
2713 	if (status & RL_ISR_SYSTEM_ERR) {
2714 		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2715 		re_init_locked(sc);
2716 	}
2717 
2718 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
2719 		if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
2720 			re_start_locked(ifp);
2721 		CSR_WRITE_2(sc, RL_IMR, intrs);
2722 	}
2723 	RL_UNLOCK(sc);
2724 }
2725 
2726 static int
2727 re_encap(struct rl_softc *sc, struct mbuf **m_head)
2728 {
2729 	struct rl_txdesc	*txd, *txd_last;
2730 	bus_dma_segment_t	segs[RL_NTXSEGS];
2731 	bus_dmamap_t		map;
2732 	struct mbuf		*m_new;
2733 	struct rl_desc		*desc;
2734 	int			nsegs, prod;
2735 	int			i, error, ei, si;
2736 	int			padlen;
2737 	uint32_t		cmdstat, csum_flags, vlanctl;
2738 
2739 	RL_LOCK_ASSERT(sc);
2740 	M_ASSERTPKTHDR((*m_head));
2741 
2742 	/*
2743 	 * With some of the RealTek chips, using the checksum offload
2744 	 * support in conjunction with the autopadding feature results
2745 	 * in the transmission of corrupt frames. For example, if we
2746 	 * need to send a really small IP fragment that's less than 60
2747 	 * bytes in size, and IP header checksumming is enabled, the
2748 	 * resulting ethernet frame that appears on the wire will
2749 	 * have garbled payload. To work around this, if TX IP checksum
2750 	 * offload is enabled, we always manually pad short frames out
2751 	 * to the minimum ethernet frame size.
2752 	 */
2753 	if ((sc->rl_flags & RL_FLAG_AUTOPAD) == 0 &&
2754 	    (*m_head)->m_pkthdr.len < RL_IP4CSUMTX_PADLEN &&
2755 	    ((*m_head)->m_pkthdr.csum_flags & CSUM_IP) != 0) {
2756 		padlen = RL_MIN_FRAMELEN - (*m_head)->m_pkthdr.len;
2757 		if (M_WRITABLE(*m_head) == 0) {
2758 			/* Get a writable copy. */
2759 			m_new = m_dup(*m_head, M_NOWAIT);
2760 			m_freem(*m_head);
2761 			if (m_new == NULL) {
2762 				*m_head = NULL;
2763 				return (ENOBUFS);
2764 			}
2765 			*m_head = m_new;
2766 		}
2767 		if ((*m_head)->m_next != NULL ||
2768 		    M_TRAILINGSPACE(*m_head) < padlen) {
2769 			m_new = m_defrag(*m_head, M_NOWAIT);
2770 			if (m_new == NULL) {
2771 				m_freem(*m_head);
2772 				*m_head = NULL;
2773 				return (ENOBUFS);
2774 			}
2775 		} else
2776 			m_new = *m_head;
2777 
2778 		/*
2779 		 * Manually pad short frames, and zero the pad space
2780 		 * to avoid leaking data.
2781 		 */
2782 		bzero(mtod(m_new, char *) + m_new->m_pkthdr.len, padlen);
2783 		m_new->m_pkthdr.len += padlen;
2784 		m_new->m_len = m_new->m_pkthdr.len;
2785 		*m_head = m_new;
2786 	}
2787 
2788 	prod = sc->rl_ldata.rl_tx_prodidx;
2789 	txd = &sc->rl_ldata.rl_tx_desc[prod];
2790 	error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap,
2791 	    *m_head, segs, &nsegs, BUS_DMA_NOWAIT);
2792 	if (error == EFBIG) {
2793 		m_new = m_collapse(*m_head, M_NOWAIT, RL_NTXSEGS);
2794 		if (m_new == NULL) {
2795 			m_freem(*m_head);
2796 			*m_head = NULL;
2797 			return (ENOBUFS);
2798 		}
2799 		*m_head = m_new;
2800 		error = bus_dmamap_load_mbuf_sg(sc->rl_ldata.rl_tx_mtag,
2801 		    txd->tx_dmamap, *m_head, segs, &nsegs, BUS_DMA_NOWAIT);
2802 		if (error != 0) {
2803 			m_freem(*m_head);
2804 			*m_head = NULL;
2805 			return (error);
2806 		}
2807 	} else if (error != 0)
2808 		return (error);
2809 	if (nsegs == 0) {
2810 		m_freem(*m_head);
2811 		*m_head = NULL;
2812 		return (EIO);
2813 	}
2814 
2815 	/* Check for number of available descriptors. */
2816 	if (sc->rl_ldata.rl_tx_free - nsegs <= 1) {
2817 		bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap);
2818 		return (ENOBUFS);
2819 	}
2820 
2821 	bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag, txd->tx_dmamap,
2822 	    BUS_DMASYNC_PREWRITE);
2823 
2824 	/*
2825 	 * Set up checksum offload. Note: checksum offload bits must
2826 	 * appear in all descriptors of a multi-descriptor transmit
2827 	 * attempt. This is according to testing done with an 8169
2828 	 * chip. This is a requirement.
2829 	 */
2830 	vlanctl = 0;
2831 	csum_flags = 0;
2832 	if (((*m_head)->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
2833 		if ((sc->rl_flags & RL_FLAG_DESCV2) != 0) {
2834 			csum_flags |= RL_TDESC_CMD_LGSEND;
2835 			vlanctl |= ((uint32_t)(*m_head)->m_pkthdr.tso_segsz <<
2836 			    RL_TDESC_CMD_MSSVALV2_SHIFT);
2837 		} else {
2838 			csum_flags |= RL_TDESC_CMD_LGSEND |
2839 			    ((uint32_t)(*m_head)->m_pkthdr.tso_segsz <<
2840 			    RL_TDESC_CMD_MSSVAL_SHIFT);
2841 		}
2842 	} else {
2843 		/*
2844 		 * Unconditionally enable IP checksum if TCP or UDP
2845 		 * checksum is required. Otherwise, TCP/UDP checksum
2846 		 * doesn't make effects.
2847 		 */
2848 		if (((*m_head)->m_pkthdr.csum_flags & RE_CSUM_FEATURES) != 0) {
2849 			if ((sc->rl_flags & RL_FLAG_DESCV2) == 0) {
2850 				csum_flags |= RL_TDESC_CMD_IPCSUM;
2851 				if (((*m_head)->m_pkthdr.csum_flags &
2852 				    CSUM_TCP) != 0)
2853 					csum_flags |= RL_TDESC_CMD_TCPCSUM;
2854 				if (((*m_head)->m_pkthdr.csum_flags &
2855 				    CSUM_UDP) != 0)
2856 					csum_flags |= RL_TDESC_CMD_UDPCSUM;
2857 			} else {
2858 				vlanctl |= RL_TDESC_CMD_IPCSUMV2;
2859 				if (((*m_head)->m_pkthdr.csum_flags &
2860 				    CSUM_TCP) != 0)
2861 					vlanctl |= RL_TDESC_CMD_TCPCSUMV2;
2862 				if (((*m_head)->m_pkthdr.csum_flags &
2863 				    CSUM_UDP) != 0)
2864 					vlanctl |= RL_TDESC_CMD_UDPCSUMV2;
2865 			}
2866 		}
2867 	}
2868 
2869 	/*
2870 	 * Set up hardware VLAN tagging. Note: vlan tag info must
2871 	 * appear in all descriptors of a multi-descriptor
2872 	 * transmission attempt.
2873 	 */
2874 	if ((*m_head)->m_flags & M_VLANTAG)
2875 		vlanctl |= bswap16((*m_head)->m_pkthdr.ether_vtag) |
2876 		    RL_TDESC_VLANCTL_TAG;
2877 
2878 	si = prod;
2879 	for (i = 0; i < nsegs; i++, prod = RL_TX_DESC_NXT(sc, prod)) {
2880 		desc = &sc->rl_ldata.rl_tx_list[prod];
2881 		desc->rl_vlanctl = htole32(vlanctl);
2882 		desc->rl_bufaddr_lo = htole32(RL_ADDR_LO(segs[i].ds_addr));
2883 		desc->rl_bufaddr_hi = htole32(RL_ADDR_HI(segs[i].ds_addr));
2884 		cmdstat = segs[i].ds_len;
2885 		if (i != 0)
2886 			cmdstat |= RL_TDESC_CMD_OWN;
2887 		if (prod == sc->rl_ldata.rl_tx_desc_cnt - 1)
2888 			cmdstat |= RL_TDESC_CMD_EOR;
2889 		desc->rl_cmdstat = htole32(cmdstat | csum_flags);
2890 		sc->rl_ldata.rl_tx_free--;
2891 	}
2892 	/* Update producer index. */
2893 	sc->rl_ldata.rl_tx_prodidx = prod;
2894 
2895 	/* Set EOF on the last descriptor. */
2896 	ei = RL_TX_DESC_PRV(sc, prod);
2897 	desc = &sc->rl_ldata.rl_tx_list[ei];
2898 	desc->rl_cmdstat |= htole32(RL_TDESC_CMD_EOF);
2899 
2900 	desc = &sc->rl_ldata.rl_tx_list[si];
2901 	/* Set SOF and transfer ownership of packet to the chip. */
2902 	desc->rl_cmdstat |= htole32(RL_TDESC_CMD_OWN | RL_TDESC_CMD_SOF);
2903 
2904 	/*
2905 	 * Insure that the map for this transmission
2906 	 * is placed at the array index of the last descriptor
2907 	 * in this chain.  (Swap last and first dmamaps.)
2908 	 */
2909 	txd_last = &sc->rl_ldata.rl_tx_desc[ei];
2910 	map = txd->tx_dmamap;
2911 	txd->tx_dmamap = txd_last->tx_dmamap;
2912 	txd_last->tx_dmamap = map;
2913 	txd_last->tx_m = *m_head;
2914 
2915 	return (0);
2916 }
2917 
2918 static void
2919 re_start(struct ifnet *ifp)
2920 {
2921 	struct rl_softc		*sc;
2922 
2923 	sc = ifp->if_softc;
2924 	RL_LOCK(sc);
2925 	re_start_locked(ifp);
2926 	RL_UNLOCK(sc);
2927 }
2928 
2929 /*
2930  * Main transmit routine for C+ and gigE NICs.
2931  */
2932 static void
2933 re_start_locked(struct ifnet *ifp)
2934 {
2935 	struct rl_softc		*sc;
2936 	struct mbuf		*m_head;
2937 	int			queued;
2938 
2939 	sc = ifp->if_softc;
2940 
2941 #ifdef DEV_NETMAP
2942 	/* XXX is this necessary ? */
2943 	if (ifp->if_capenable & IFCAP_NETMAP) {
2944 		struct netmap_kring *kring = NA(ifp)->tx_rings[0];
2945 		if (sc->rl_ldata.rl_tx_prodidx != kring->nr_hwcur) {
2946 			/* kick the tx unit */
2947 			CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
2948 #ifdef RE_TX_MODERATION
2949 			CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2950 #endif
2951 			sc->rl_watchdog_timer = 5;
2952 		}
2953 		return;
2954 	}
2955 #endif /* DEV_NETMAP */
2956 
2957 	if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
2958 	    IFF_DRV_RUNNING || (sc->rl_flags & RL_FLAG_LINK) == 0)
2959 		return;
2960 
2961 	for (queued = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd) &&
2962 	    sc->rl_ldata.rl_tx_free > 1;) {
2963 		IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
2964 		if (m_head == NULL)
2965 			break;
2966 
2967 		if (re_encap(sc, &m_head) != 0) {
2968 			if (m_head == NULL)
2969 				break;
2970 			IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
2971 			ifp->if_drv_flags |= IFF_DRV_OACTIVE;
2972 			break;
2973 		}
2974 
2975 		/*
2976 		 * If there's a BPF listener, bounce a copy of this frame
2977 		 * to him.
2978 		 */
2979 		ETHER_BPF_MTAP(ifp, m_head);
2980 
2981 		queued++;
2982 	}
2983 
2984 	if (queued == 0) {
2985 #ifdef RE_TX_MODERATION
2986 		if (sc->rl_ldata.rl_tx_free != sc->rl_ldata.rl_tx_desc_cnt)
2987 			CSR_WRITE_4(sc, RL_TIMERCNT, 1);
2988 #endif
2989 		return;
2990 	}
2991 
2992 	re_start_tx(sc);
2993 }
2994 
2995 static void
2996 re_start_tx(struct rl_softc *sc)
2997 {
2998 
2999 	/* Flush the TX descriptors */
3000 	bus_dmamap_sync(sc->rl_ldata.rl_tx_list_tag,
3001 	    sc->rl_ldata.rl_tx_list_map,
3002 	    BUS_DMASYNC_PREWRITE|BUS_DMASYNC_PREREAD);
3003 
3004 	CSR_WRITE_1(sc, sc->rl_txstart, RL_TXSTART_START);
3005 
3006 #ifdef RE_TX_MODERATION
3007 	/*
3008 	 * Use the countdown timer for interrupt moderation.
3009 	 * 'TX done' interrupts are disabled. Instead, we reset the
3010 	 * countdown timer, which will begin counting until it hits
3011 	 * the value in the TIMERINT register, and then trigger an
3012 	 * interrupt. Each time we write to the TIMERCNT register,
3013 	 * the timer count is reset to 0.
3014 	 */
3015 	CSR_WRITE_4(sc, RL_TIMERCNT, 1);
3016 #endif
3017 
3018 	/*
3019 	 * Set a timeout in case the chip goes out to lunch.
3020 	 */
3021 	sc->rl_watchdog_timer = 5;
3022 }
3023 
3024 static void
3025 re_set_jumbo(struct rl_softc *sc, int jumbo)
3026 {
3027 
3028 	if (sc->rl_hwrev->rl_rev == RL_HWREV_8168E_VL) {
3029 		pci_set_max_read_req(sc->rl_dev, 4096);
3030 		return;
3031 	}
3032 
3033 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_WRITECFG);
3034 	if (jumbo != 0) {
3035 		CSR_WRITE_1(sc, sc->rl_cfg3, CSR_READ_1(sc, sc->rl_cfg3) |
3036 		    RL_CFG3_JUMBO_EN0);
3037 		switch (sc->rl_hwrev->rl_rev) {
3038 		case RL_HWREV_8168DP:
3039 			break;
3040 		case RL_HWREV_8168E:
3041 			CSR_WRITE_1(sc, sc->rl_cfg4,
3042 			    CSR_READ_1(sc, sc->rl_cfg4) | 0x01);
3043 			break;
3044 		default:
3045 			CSR_WRITE_1(sc, sc->rl_cfg4,
3046 			    CSR_READ_1(sc, sc->rl_cfg4) | RL_CFG4_JUMBO_EN1);
3047 		}
3048 	} else {
3049 		CSR_WRITE_1(sc, sc->rl_cfg3, CSR_READ_1(sc, sc->rl_cfg3) &
3050 		    ~RL_CFG3_JUMBO_EN0);
3051 		switch (sc->rl_hwrev->rl_rev) {
3052 		case RL_HWREV_8168DP:
3053 			break;
3054 		case RL_HWREV_8168E:
3055 			CSR_WRITE_1(sc, sc->rl_cfg4,
3056 			    CSR_READ_1(sc, sc->rl_cfg4) & ~0x01);
3057 			break;
3058 		default:
3059 			CSR_WRITE_1(sc, sc->rl_cfg4,
3060 			    CSR_READ_1(sc, sc->rl_cfg4) & ~RL_CFG4_JUMBO_EN1);
3061 		}
3062 	}
3063 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
3064 
3065 	switch (sc->rl_hwrev->rl_rev) {
3066 	case RL_HWREV_8168DP:
3067 		pci_set_max_read_req(sc->rl_dev, 4096);
3068 		break;
3069 	default:
3070 		if (jumbo != 0)
3071 			pci_set_max_read_req(sc->rl_dev, 512);
3072 		else
3073 			pci_set_max_read_req(sc->rl_dev, 4096);
3074 	}
3075 }
3076 
3077 static void
3078 re_init(void *xsc)
3079 {
3080 	struct rl_softc		*sc = xsc;
3081 
3082 	RL_LOCK(sc);
3083 	re_init_locked(sc);
3084 	RL_UNLOCK(sc);
3085 }
3086 
3087 static void
3088 re_init_locked(struct rl_softc *sc)
3089 {
3090 	struct ifnet		*ifp = sc->rl_ifp;
3091 	struct mii_data		*mii;
3092 	uint32_t		reg;
3093 	uint16_t		cfg;
3094 	union {
3095 		uint32_t align_dummy;
3096 		u_char eaddr[ETHER_ADDR_LEN];
3097         } eaddr;
3098 
3099 	RL_LOCK_ASSERT(sc);
3100 
3101 	mii = device_get_softc(sc->rl_miibus);
3102 
3103 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
3104 		return;
3105 
3106 	/*
3107 	 * Cancel pending I/O and free all RX/TX buffers.
3108 	 */
3109 	re_stop(sc);
3110 
3111 	/* Put controller into known state. */
3112 	re_reset(sc);
3113 
3114 	/*
3115 	 * For C+ mode, initialize the RX descriptors and mbufs.
3116 	 */
3117 	if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) {
3118 		if (ifp->if_mtu > RL_MTU) {
3119 			if (re_jrx_list_init(sc) != 0) {
3120 				device_printf(sc->rl_dev,
3121 				    "no memory for jumbo RX buffers\n");
3122 				re_stop(sc);
3123 				return;
3124 			}
3125 			/* Disable checksum offloading for jumbo frames. */
3126 			ifp->if_capenable &= ~(IFCAP_HWCSUM | IFCAP_TSO4);
3127 			ifp->if_hwassist &= ~(RE_CSUM_FEATURES | CSUM_TSO);
3128 		} else {
3129 			if (re_rx_list_init(sc) != 0) {
3130 				device_printf(sc->rl_dev,
3131 				    "no memory for RX buffers\n");
3132 				re_stop(sc);
3133 				return;
3134 			}
3135 		}
3136 		re_set_jumbo(sc, ifp->if_mtu > RL_MTU);
3137 	} else {
3138 		if (re_rx_list_init(sc) != 0) {
3139 			device_printf(sc->rl_dev, "no memory for RX buffers\n");
3140 			re_stop(sc);
3141 			return;
3142 		}
3143 		if ((sc->rl_flags & RL_FLAG_PCIE) != 0 &&
3144 		    pci_get_device(sc->rl_dev) != RT_DEVICEID_8101E) {
3145 			if (ifp->if_mtu > RL_MTU)
3146 				pci_set_max_read_req(sc->rl_dev, 512);
3147 			else
3148 				pci_set_max_read_req(sc->rl_dev, 4096);
3149 		}
3150 	}
3151 	re_tx_list_init(sc);
3152 
3153 	/*
3154 	 * Enable C+ RX and TX mode, as well as VLAN stripping and
3155 	 * RX checksum offload. We must configure the C+ register
3156 	 * before all others.
3157 	 */
3158 	cfg = RL_CPLUSCMD_PCI_MRW;
3159 	if ((ifp->if_capenable & IFCAP_RXCSUM) != 0)
3160 		cfg |= RL_CPLUSCMD_RXCSUM_ENB;
3161 	if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0)
3162 		cfg |= RL_CPLUSCMD_VLANSTRIP;
3163 	if ((sc->rl_flags & RL_FLAG_MACSTAT) != 0) {
3164 		cfg |= RL_CPLUSCMD_MACSTAT_DIS;
3165 		/* XXX magic. */
3166 		cfg |= 0x0001;
3167 	} else
3168 		cfg |= RL_CPLUSCMD_RXENB | RL_CPLUSCMD_TXENB;
3169 	CSR_WRITE_2(sc, RL_CPLUS_CMD, cfg);
3170 	if (sc->rl_hwrev->rl_rev == RL_HWREV_8169_8110SC ||
3171 	    sc->rl_hwrev->rl_rev == RL_HWREV_8169_8110SCE) {
3172 		reg = 0x000fff00;
3173 		if ((CSR_READ_1(sc, sc->rl_cfg2) & RL_CFG2_PCI66MHZ) != 0)
3174 			reg |= 0x000000ff;
3175 		if (sc->rl_hwrev->rl_rev == RL_HWREV_8169_8110SCE)
3176 			reg |= 0x00f00000;
3177 		CSR_WRITE_4(sc, 0x7c, reg);
3178 		/* Disable interrupt mitigation. */
3179 		CSR_WRITE_2(sc, 0xe2, 0);
3180 	}
3181 	/*
3182 	 * Disable TSO if interface MTU size is greater than MSS
3183 	 * allowed in controller.
3184 	 */
3185 	if (ifp->if_mtu > RL_TSO_MTU && (ifp->if_capenable & IFCAP_TSO4) != 0) {
3186 		ifp->if_capenable &= ~IFCAP_TSO4;
3187 		ifp->if_hwassist &= ~CSUM_TSO;
3188 	}
3189 
3190 	/*
3191 	 * Init our MAC address.  Even though the chipset
3192 	 * documentation doesn't mention it, we need to enter "Config
3193 	 * register write enable" mode to modify the ID registers.
3194 	 */
3195 	/* Copy MAC address on stack to align. */
3196 	bcopy(IF_LLADDR(ifp), eaddr.eaddr, ETHER_ADDR_LEN);
3197 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_WRITECFG);
3198 	CSR_WRITE_4(sc, RL_IDR0,
3199 	    htole32(*(u_int32_t *)(&eaddr.eaddr[0])));
3200 	CSR_WRITE_4(sc, RL_IDR4,
3201 	    htole32(*(u_int32_t *)(&eaddr.eaddr[4])));
3202 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
3203 
3204 	/*
3205 	 * Load the addresses of the RX and TX lists into the chip.
3206 	 */
3207 
3208 	CSR_WRITE_4(sc, RL_RXLIST_ADDR_HI,
3209 	    RL_ADDR_HI(sc->rl_ldata.rl_rx_list_addr));
3210 	CSR_WRITE_4(sc, RL_RXLIST_ADDR_LO,
3211 	    RL_ADDR_LO(sc->rl_ldata.rl_rx_list_addr));
3212 
3213 	CSR_WRITE_4(sc, RL_TXLIST_ADDR_HI,
3214 	    RL_ADDR_HI(sc->rl_ldata.rl_tx_list_addr));
3215 	CSR_WRITE_4(sc, RL_TXLIST_ADDR_LO,
3216 	    RL_ADDR_LO(sc->rl_ldata.rl_tx_list_addr));
3217 
3218 	if ((sc->rl_flags & RL_FLAG_8168G_PLUS) != 0) {
3219 		/* Disable RXDV gate. */
3220 		CSR_WRITE_4(sc, RL_MISC, CSR_READ_4(sc, RL_MISC) &
3221 		    ~0x00080000);
3222 	}
3223 
3224 	/*
3225 	 * Enable transmit and receive for pre-RTL8168G controllers.
3226 	 * RX/TX MACs should be enabled before RX/TX configuration.
3227 	 */
3228 	if ((sc->rl_flags & RL_FLAG_8168G_PLUS) == 0)
3229 		CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_TX_ENB | RL_CMD_RX_ENB);
3230 
3231 	/*
3232 	 * Set the initial TX configuration.
3233 	 */
3234 	if (sc->rl_testmode) {
3235 		if (sc->rl_type == RL_8169)
3236 			CSR_WRITE_4(sc, RL_TXCFG,
3237 			    RL_TXCFG_CONFIG|RL_LOOPTEST_ON);
3238 		else
3239 			CSR_WRITE_4(sc, RL_TXCFG,
3240 			    RL_TXCFG_CONFIG|RL_LOOPTEST_ON_CPLUS);
3241 	} else
3242 		CSR_WRITE_4(sc, RL_TXCFG, RL_TXCFG_CONFIG);
3243 
3244 	CSR_WRITE_1(sc, RL_EARLY_TX_THRESH, 16);
3245 
3246 	/*
3247 	 * Set the initial RX configuration.
3248 	 */
3249 	re_set_rxmode(sc);
3250 
3251 	/* Configure interrupt moderation. */
3252 	if (sc->rl_type == RL_8169) {
3253 		/* Magic from vendor. */
3254 		CSR_WRITE_2(sc, RL_INTRMOD, 0x5100);
3255 	}
3256 
3257 	/*
3258 	 * Enable transmit and receive for RTL8168G and later controllers.
3259 	 * RX/TX MACs should be enabled after RX/TX configuration.
3260 	 */
3261 	if ((sc->rl_flags & RL_FLAG_8168G_PLUS) != 0)
3262 		CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_TX_ENB | RL_CMD_RX_ENB);
3263 
3264 #ifdef DEVICE_POLLING
3265 	/*
3266 	 * Disable interrupts if we are polling.
3267 	 */
3268 	if (ifp->if_capenable & IFCAP_POLLING)
3269 		CSR_WRITE_2(sc, RL_IMR, 0);
3270 	else	/* otherwise ... */
3271 #endif
3272 
3273 	/*
3274 	 * Enable interrupts.
3275 	 */
3276 	if (sc->rl_testmode)
3277 		CSR_WRITE_2(sc, RL_IMR, 0);
3278 	else
3279 		CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS);
3280 	CSR_WRITE_2(sc, RL_ISR, RL_INTRS_CPLUS);
3281 
3282 	/* Set initial TX threshold */
3283 	sc->rl_txthresh = RL_TX_THRESH_INIT;
3284 
3285 	/* Start RX/TX process. */
3286 	CSR_WRITE_4(sc, RL_MISSEDPKT, 0);
3287 
3288 	/*
3289 	 * Initialize the timer interrupt register so that
3290 	 * a timer interrupt will be generated once the timer
3291 	 * reaches a certain number of ticks. The timer is
3292 	 * reloaded on each transmit.
3293 	 */
3294 #ifdef RE_TX_MODERATION
3295 	/*
3296 	 * Use timer interrupt register to moderate TX interrupt
3297 	 * moderation, which dramatically improves TX frame rate.
3298 	 */
3299 	if (sc->rl_type == RL_8169)
3300 		CSR_WRITE_4(sc, RL_TIMERINT_8169, 0x800);
3301 	else
3302 		CSR_WRITE_4(sc, RL_TIMERINT, 0x400);
3303 #else
3304 	/*
3305 	 * Use timer interrupt register to moderate RX interrupt
3306 	 * moderation.
3307 	 */
3308 	if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) != 0 &&
3309 	    intr_filter == 0) {
3310 		if (sc->rl_type == RL_8169)
3311 			CSR_WRITE_4(sc, RL_TIMERINT_8169,
3312 			    RL_USECS(sc->rl_int_rx_mod));
3313 	} else {
3314 		if (sc->rl_type == RL_8169)
3315 			CSR_WRITE_4(sc, RL_TIMERINT_8169, RL_USECS(0));
3316 	}
3317 #endif
3318 
3319 	/*
3320 	 * For 8169 gigE NICs, set the max allowed RX packet
3321 	 * size so we can receive jumbo frames.
3322 	 */
3323 	if (sc->rl_type == RL_8169) {
3324 		if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) {
3325 			/*
3326 			 * For controllers that use new jumbo frame scheme,
3327 			 * set maximum size of jumbo frame depending on
3328 			 * controller revisions.
3329 			 */
3330 			if (ifp->if_mtu > RL_MTU)
3331 				CSR_WRITE_2(sc, RL_MAXRXPKTLEN,
3332 				    sc->rl_hwrev->rl_max_mtu +
3333 				    ETHER_VLAN_ENCAP_LEN + ETHER_HDR_LEN +
3334 				    ETHER_CRC_LEN);
3335 			else
3336 				CSR_WRITE_2(sc, RL_MAXRXPKTLEN,
3337 				    RE_RX_DESC_BUFLEN);
3338 		} else if ((sc->rl_flags & RL_FLAG_PCIE) != 0 &&
3339 		    sc->rl_hwrev->rl_max_mtu == RL_MTU) {
3340 			/* RTL810x has no jumbo frame support. */
3341 			CSR_WRITE_2(sc, RL_MAXRXPKTLEN, RE_RX_DESC_BUFLEN);
3342 		} else
3343 			CSR_WRITE_2(sc, RL_MAXRXPKTLEN, 16383);
3344 	}
3345 
3346 	if (sc->rl_testmode)
3347 		return;
3348 
3349 	CSR_WRITE_1(sc, sc->rl_cfg1, CSR_READ_1(sc, sc->rl_cfg1) |
3350 	    RL_CFG1_DRVLOAD);
3351 
3352 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
3353 	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
3354 
3355 	sc->rl_flags &= ~RL_FLAG_LINK;
3356 	mii_mediachg(mii);
3357 
3358 	sc->rl_watchdog_timer = 0;
3359 	callout_reset(&sc->rl_stat_callout, hz, re_tick, sc);
3360 }
3361 
3362 /*
3363  * Set media options.
3364  */
3365 static int
3366 re_ifmedia_upd(struct ifnet *ifp)
3367 {
3368 	struct rl_softc		*sc;
3369 	struct mii_data		*mii;
3370 	int			error;
3371 
3372 	sc = ifp->if_softc;
3373 	mii = device_get_softc(sc->rl_miibus);
3374 	RL_LOCK(sc);
3375 	error = mii_mediachg(mii);
3376 	RL_UNLOCK(sc);
3377 
3378 	return (error);
3379 }
3380 
3381 /*
3382  * Report current media status.
3383  */
3384 static void
3385 re_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
3386 {
3387 	struct rl_softc		*sc;
3388 	struct mii_data		*mii;
3389 
3390 	sc = ifp->if_softc;
3391 	mii = device_get_softc(sc->rl_miibus);
3392 
3393 	RL_LOCK(sc);
3394 	mii_pollstat(mii);
3395 	ifmr->ifm_active = mii->mii_media_active;
3396 	ifmr->ifm_status = mii->mii_media_status;
3397 	RL_UNLOCK(sc);
3398 }
3399 
3400 static int
3401 re_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
3402 {
3403 	struct rl_softc		*sc = ifp->if_softc;
3404 	struct ifreq		*ifr = (struct ifreq *) data;
3405 	struct mii_data		*mii;
3406 	int			error = 0;
3407 
3408 	switch (command) {
3409 	case SIOCSIFMTU:
3410 		if (ifr->ifr_mtu < ETHERMIN ||
3411 		    ifr->ifr_mtu > sc->rl_hwrev->rl_max_mtu ||
3412 		    ((sc->rl_flags & RL_FLAG_FASTETHER) != 0 &&
3413 		    ifr->ifr_mtu > RL_MTU)) {
3414 			error = EINVAL;
3415 			break;
3416 		}
3417 		RL_LOCK(sc);
3418 		if (ifp->if_mtu != ifr->ifr_mtu) {
3419 			ifp->if_mtu = ifr->ifr_mtu;
3420 			if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0 &&
3421 			    (ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
3422 				ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
3423 				re_init_locked(sc);
3424 			}
3425 			if (ifp->if_mtu > RL_TSO_MTU &&
3426 			    (ifp->if_capenable & IFCAP_TSO4) != 0) {
3427 				ifp->if_capenable &= ~(IFCAP_TSO4 |
3428 				    IFCAP_VLAN_HWTSO);
3429 				ifp->if_hwassist &= ~CSUM_TSO;
3430 			}
3431 			VLAN_CAPABILITIES(ifp);
3432 		}
3433 		RL_UNLOCK(sc);
3434 		break;
3435 	case SIOCSIFFLAGS:
3436 		RL_LOCK(sc);
3437 		if ((ifp->if_flags & IFF_UP) != 0) {
3438 			if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
3439 				if (((ifp->if_flags ^ sc->rl_if_flags)
3440 				    & (IFF_PROMISC | IFF_ALLMULTI)) != 0)
3441 					re_set_rxmode(sc);
3442 			} else
3443 				re_init_locked(sc);
3444 		} else {
3445 			if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
3446 				re_stop(sc);
3447 		}
3448 		sc->rl_if_flags = ifp->if_flags;
3449 		RL_UNLOCK(sc);
3450 		break;
3451 	case SIOCADDMULTI:
3452 	case SIOCDELMULTI:
3453 		RL_LOCK(sc);
3454 		if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
3455 			re_set_rxmode(sc);
3456 		RL_UNLOCK(sc);
3457 		break;
3458 	case SIOCGIFMEDIA:
3459 	case SIOCSIFMEDIA:
3460 		mii = device_get_softc(sc->rl_miibus);
3461 		error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
3462 		break;
3463 	case SIOCSIFCAP:
3464 	    {
3465 		int mask, reinit;
3466 
3467 		mask = ifr->ifr_reqcap ^ ifp->if_capenable;
3468 		reinit = 0;
3469 #ifdef DEVICE_POLLING
3470 		if (mask & IFCAP_POLLING) {
3471 			if (ifr->ifr_reqcap & IFCAP_POLLING) {
3472 				error = ether_poll_register(re_poll, ifp);
3473 				if (error)
3474 					return (error);
3475 				RL_LOCK(sc);
3476 				/* Disable interrupts */
3477 				CSR_WRITE_2(sc, RL_IMR, 0x0000);
3478 				ifp->if_capenable |= IFCAP_POLLING;
3479 				RL_UNLOCK(sc);
3480 			} else {
3481 				error = ether_poll_deregister(ifp);
3482 				/* Enable interrupts. */
3483 				RL_LOCK(sc);
3484 				CSR_WRITE_2(sc, RL_IMR, RL_INTRS_CPLUS);
3485 				ifp->if_capenable &= ~IFCAP_POLLING;
3486 				RL_UNLOCK(sc);
3487 			}
3488 		}
3489 #endif /* DEVICE_POLLING */
3490 		RL_LOCK(sc);
3491 		if ((mask & IFCAP_TXCSUM) != 0 &&
3492 		    (ifp->if_capabilities & IFCAP_TXCSUM) != 0) {
3493 			ifp->if_capenable ^= IFCAP_TXCSUM;
3494 			if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
3495 				ifp->if_hwassist |= RE_CSUM_FEATURES;
3496 			else
3497 				ifp->if_hwassist &= ~RE_CSUM_FEATURES;
3498 			reinit = 1;
3499 		}
3500 		if ((mask & IFCAP_RXCSUM) != 0 &&
3501 		    (ifp->if_capabilities & IFCAP_RXCSUM) != 0) {
3502 			ifp->if_capenable ^= IFCAP_RXCSUM;
3503 			reinit = 1;
3504 		}
3505 		if ((mask & IFCAP_TSO4) != 0 &&
3506 		    (ifp->if_capabilities & IFCAP_TSO4) != 0) {
3507 			ifp->if_capenable ^= IFCAP_TSO4;
3508 			if ((IFCAP_TSO4 & ifp->if_capenable) != 0)
3509 				ifp->if_hwassist |= CSUM_TSO;
3510 			else
3511 				ifp->if_hwassist &= ~CSUM_TSO;
3512 			if (ifp->if_mtu > RL_TSO_MTU &&
3513 			    (ifp->if_capenable & IFCAP_TSO4) != 0) {
3514 				ifp->if_capenable &= ~IFCAP_TSO4;
3515 				ifp->if_hwassist &= ~CSUM_TSO;
3516 			}
3517 		}
3518 		if ((mask & IFCAP_VLAN_HWTSO) != 0 &&
3519 		    (ifp->if_capabilities & IFCAP_VLAN_HWTSO) != 0)
3520 			ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
3521 		if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
3522 		    (ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0) {
3523 			ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
3524 			/* TSO over VLAN requires VLAN hardware tagging. */
3525 			if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0)
3526 				ifp->if_capenable &= ~IFCAP_VLAN_HWTSO;
3527 			reinit = 1;
3528 		}
3529 		if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0 &&
3530 		    (mask & (IFCAP_HWCSUM | IFCAP_TSO4 |
3531 		    IFCAP_VLAN_HWTSO)) != 0)
3532 				reinit = 1;
3533 		if ((mask & IFCAP_WOL) != 0 &&
3534 		    (ifp->if_capabilities & IFCAP_WOL) != 0) {
3535 			if ((mask & IFCAP_WOL_UCAST) != 0)
3536 				ifp->if_capenable ^= IFCAP_WOL_UCAST;
3537 			if ((mask & IFCAP_WOL_MCAST) != 0)
3538 				ifp->if_capenable ^= IFCAP_WOL_MCAST;
3539 			if ((mask & IFCAP_WOL_MAGIC) != 0)
3540 				ifp->if_capenable ^= IFCAP_WOL_MAGIC;
3541 		}
3542 		if (reinit && ifp->if_drv_flags & IFF_DRV_RUNNING) {
3543 			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
3544 			re_init_locked(sc);
3545 		}
3546 		RL_UNLOCK(sc);
3547 		VLAN_CAPABILITIES(ifp);
3548 	    }
3549 		break;
3550 	default:
3551 		error = ether_ioctl(ifp, command, data);
3552 		break;
3553 	}
3554 
3555 	return (error);
3556 }
3557 
3558 static void
3559 re_watchdog(struct rl_softc *sc)
3560 {
3561 	struct ifnet		*ifp;
3562 
3563 	RL_LOCK_ASSERT(sc);
3564 
3565 	if (sc->rl_watchdog_timer == 0 || --sc->rl_watchdog_timer != 0)
3566 		return;
3567 
3568 	ifp = sc->rl_ifp;
3569 	re_txeof(sc);
3570 	if (sc->rl_ldata.rl_tx_free == sc->rl_ldata.rl_tx_desc_cnt) {
3571 		if_printf(ifp, "watchdog timeout (missed Tx interrupts) "
3572 		    "-- recovering\n");
3573 		if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
3574 			re_start_locked(ifp);
3575 		return;
3576 	}
3577 
3578 	if_printf(ifp, "watchdog timeout\n");
3579 	if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
3580 
3581 	re_rxeof(sc, NULL);
3582 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
3583 	re_init_locked(sc);
3584 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
3585 		re_start_locked(ifp);
3586 }
3587 
3588 /*
3589  * Stop the adapter and free any mbufs allocated to the
3590  * RX and TX lists.
3591  */
3592 static void
3593 re_stop(struct rl_softc *sc)
3594 {
3595 	int			i;
3596 	struct ifnet		*ifp;
3597 	struct rl_txdesc	*txd;
3598 	struct rl_rxdesc	*rxd;
3599 
3600 	RL_LOCK_ASSERT(sc);
3601 
3602 	ifp = sc->rl_ifp;
3603 
3604 	sc->rl_watchdog_timer = 0;
3605 	callout_stop(&sc->rl_stat_callout);
3606 	ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
3607 
3608 	/*
3609 	 * Disable accepting frames to put RX MAC into idle state.
3610 	 * Otherwise it's possible to get frames while stop command
3611 	 * execution is in progress and controller can DMA the frame
3612 	 * to already freed RX buffer during that period.
3613 	 */
3614 	CSR_WRITE_4(sc, RL_RXCFG, CSR_READ_4(sc, RL_RXCFG) &
3615 	    ~(RL_RXCFG_RX_ALLPHYS | RL_RXCFG_RX_INDIV | RL_RXCFG_RX_MULTI |
3616 	    RL_RXCFG_RX_BROAD));
3617 
3618 	if ((sc->rl_flags & RL_FLAG_8168G_PLUS) != 0) {
3619 		/* Enable RXDV gate. */
3620 		CSR_WRITE_4(sc, RL_MISC, CSR_READ_4(sc, RL_MISC) |
3621 		    0x00080000);
3622 	}
3623 
3624 	if ((sc->rl_flags & RL_FLAG_WAIT_TXPOLL) != 0) {
3625 		for (i = RL_TIMEOUT; i > 0; i--) {
3626 			if ((CSR_READ_1(sc, sc->rl_txstart) &
3627 			    RL_TXSTART_START) == 0)
3628 				break;
3629 			DELAY(20);
3630 		}
3631 		if (i == 0)
3632 			device_printf(sc->rl_dev,
3633 			    "stopping TX poll timed out!\n");
3634 		CSR_WRITE_1(sc, RL_COMMAND, 0x00);
3635 	} else if ((sc->rl_flags & RL_FLAG_CMDSTOP) != 0) {
3636 		CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_STOPREQ | RL_CMD_TX_ENB |
3637 		    RL_CMD_RX_ENB);
3638 		if ((sc->rl_flags & RL_FLAG_CMDSTOP_WAIT_TXQ) != 0) {
3639 			for (i = RL_TIMEOUT; i > 0; i--) {
3640 				if ((CSR_READ_4(sc, RL_TXCFG) &
3641 				    RL_TXCFG_QUEUE_EMPTY) != 0)
3642 					break;
3643 				DELAY(100);
3644 			}
3645 			if (i == 0)
3646 				device_printf(sc->rl_dev,
3647 				   "stopping TXQ timed out!\n");
3648 		}
3649 	} else
3650 		CSR_WRITE_1(sc, RL_COMMAND, 0x00);
3651 	DELAY(1000);
3652 	CSR_WRITE_2(sc, RL_IMR, 0x0000);
3653 	CSR_WRITE_2(sc, RL_ISR, 0xFFFF);
3654 
3655 	if (sc->rl_head != NULL) {
3656 		m_freem(sc->rl_head);
3657 		sc->rl_head = sc->rl_tail = NULL;
3658 	}
3659 
3660 	/* Free the TX list buffers. */
3661 	for (i = 0; i < sc->rl_ldata.rl_tx_desc_cnt; i++) {
3662 		txd = &sc->rl_ldata.rl_tx_desc[i];
3663 		if (txd->tx_m != NULL) {
3664 			bus_dmamap_sync(sc->rl_ldata.rl_tx_mtag,
3665 			    txd->tx_dmamap, BUS_DMASYNC_POSTWRITE);
3666 			bus_dmamap_unload(sc->rl_ldata.rl_tx_mtag,
3667 			    txd->tx_dmamap);
3668 			m_freem(txd->tx_m);
3669 			txd->tx_m = NULL;
3670 		}
3671 	}
3672 
3673 	/* Free the RX list buffers. */
3674 	for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
3675 		rxd = &sc->rl_ldata.rl_rx_desc[i];
3676 		if (rxd->rx_m != NULL) {
3677 			bus_dmamap_sync(sc->rl_ldata.rl_rx_mtag,
3678 			    rxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
3679 			bus_dmamap_unload(sc->rl_ldata.rl_rx_mtag,
3680 			    rxd->rx_dmamap);
3681 			m_freem(rxd->rx_m);
3682 			rxd->rx_m = NULL;
3683 		}
3684 	}
3685 
3686 	if ((sc->rl_flags & RL_FLAG_JUMBOV2) != 0) {
3687 		for (i = 0; i < sc->rl_ldata.rl_rx_desc_cnt; i++) {
3688 			rxd = &sc->rl_ldata.rl_jrx_desc[i];
3689 			if (rxd->rx_m != NULL) {
3690 				bus_dmamap_sync(sc->rl_ldata.rl_jrx_mtag,
3691 				    rxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
3692 				bus_dmamap_unload(sc->rl_ldata.rl_jrx_mtag,
3693 				    rxd->rx_dmamap);
3694 				m_freem(rxd->rx_m);
3695 				rxd->rx_m = NULL;
3696 			}
3697 		}
3698 	}
3699 }
3700 
3701 /*
3702  * Device suspend routine.  Stop the interface and save some PCI
3703  * settings in case the BIOS doesn't restore them properly on
3704  * resume.
3705  */
3706 static int
3707 re_suspend(device_t dev)
3708 {
3709 	struct rl_softc		*sc;
3710 
3711 	sc = device_get_softc(dev);
3712 
3713 	RL_LOCK(sc);
3714 	re_stop(sc);
3715 	re_setwol(sc);
3716 	sc->suspended = 1;
3717 	RL_UNLOCK(sc);
3718 
3719 	return (0);
3720 }
3721 
3722 /*
3723  * Device resume routine.  Restore some PCI settings in case the BIOS
3724  * doesn't, re-enable busmastering, and restart the interface if
3725  * appropriate.
3726  */
3727 static int
3728 re_resume(device_t dev)
3729 {
3730 	struct rl_softc		*sc;
3731 	struct ifnet		*ifp;
3732 
3733 	sc = device_get_softc(dev);
3734 
3735 	RL_LOCK(sc);
3736 
3737 	ifp = sc->rl_ifp;
3738 	/* Take controller out of sleep mode. */
3739 	if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) {
3740 		if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80)
3741 			CSR_WRITE_1(sc, RL_GPIO,
3742 			    CSR_READ_1(sc, RL_GPIO) | 0x01);
3743 	}
3744 
3745 	/*
3746 	 * Clear WOL matching such that normal Rx filtering
3747 	 * wouldn't interfere with WOL patterns.
3748 	 */
3749 	re_clrwol(sc);
3750 
3751 	/* reinitialize interface if necessary */
3752 	if (ifp->if_flags & IFF_UP)
3753 		re_init_locked(sc);
3754 
3755 	sc->suspended = 0;
3756 	RL_UNLOCK(sc);
3757 
3758 	return (0);
3759 }
3760 
3761 /*
3762  * Stop all chip I/O so that the kernel's probe routines don't
3763  * get confused by errant DMAs when rebooting.
3764  */
3765 static int
3766 re_shutdown(device_t dev)
3767 {
3768 	struct rl_softc		*sc;
3769 
3770 	sc = device_get_softc(dev);
3771 
3772 	RL_LOCK(sc);
3773 	re_stop(sc);
3774 	/*
3775 	 * Mark interface as down since otherwise we will panic if
3776 	 * interrupt comes in later on, which can happen in some
3777 	 * cases.
3778 	 */
3779 	sc->rl_ifp->if_flags &= ~IFF_UP;
3780 	re_setwol(sc);
3781 	RL_UNLOCK(sc);
3782 
3783 	return (0);
3784 }
3785 
3786 static void
3787 re_set_linkspeed(struct rl_softc *sc)
3788 {
3789 	struct mii_softc *miisc;
3790 	struct mii_data *mii;
3791 	int aneg, i, phyno;
3792 
3793 	RL_LOCK_ASSERT(sc);
3794 
3795 	mii = device_get_softc(sc->rl_miibus);
3796 	mii_pollstat(mii);
3797 	aneg = 0;
3798 	if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
3799 	    (IFM_ACTIVE | IFM_AVALID)) {
3800 		switch IFM_SUBTYPE(mii->mii_media_active) {
3801 		case IFM_10_T:
3802 		case IFM_100_TX:
3803 			return;
3804 		case IFM_1000_T:
3805 			aneg++;
3806 			break;
3807 		default:
3808 			break;
3809 		}
3810 	}
3811 	miisc = LIST_FIRST(&mii->mii_phys);
3812 	phyno = miisc->mii_phy;
3813 	LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
3814 		PHY_RESET(miisc);
3815 	re_miibus_writereg(sc->rl_dev, phyno, MII_100T2CR, 0);
3816 	re_miibus_writereg(sc->rl_dev, phyno,
3817 	    MII_ANAR, ANAR_TX_FD | ANAR_TX | ANAR_10_FD | ANAR_10 | ANAR_CSMA);
3818 	re_miibus_writereg(sc->rl_dev, phyno,
3819 	    MII_BMCR, BMCR_AUTOEN | BMCR_STARTNEG);
3820 	DELAY(1000);
3821 	if (aneg != 0) {
3822 		/*
3823 		 * Poll link state until re(4) get a 10/100Mbps link.
3824 		 */
3825 		for (i = 0; i < MII_ANEGTICKS_GIGE; i++) {
3826 			mii_pollstat(mii);
3827 			if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID))
3828 			    == (IFM_ACTIVE | IFM_AVALID)) {
3829 				switch (IFM_SUBTYPE(mii->mii_media_active)) {
3830 				case IFM_10_T:
3831 				case IFM_100_TX:
3832 					return;
3833 				default:
3834 					break;
3835 				}
3836 			}
3837 			RL_UNLOCK(sc);
3838 			pause("relnk", hz);
3839 			RL_LOCK(sc);
3840 		}
3841 		if (i == MII_ANEGTICKS_GIGE)
3842 			device_printf(sc->rl_dev,
3843 			    "establishing a link failed, WOL may not work!");
3844 	}
3845 	/*
3846 	 * No link, force MAC to have 100Mbps, full-duplex link.
3847 	 * MAC does not require reprogramming on resolved speed/duplex,
3848 	 * so this is just for completeness.
3849 	 */
3850 	mii->mii_media_status = IFM_AVALID | IFM_ACTIVE;
3851 	mii->mii_media_active = IFM_ETHER | IFM_100_TX | IFM_FDX;
3852 }
3853 
3854 static void
3855 re_setwol(struct rl_softc *sc)
3856 {
3857 	struct ifnet		*ifp;
3858 	int			pmc;
3859 	uint16_t		pmstat;
3860 	uint8_t			v;
3861 
3862 	RL_LOCK_ASSERT(sc);
3863 
3864 	if (pci_find_cap(sc->rl_dev, PCIY_PMG, &pmc) != 0)
3865 		return;
3866 
3867 	ifp = sc->rl_ifp;
3868 	/* Put controller into sleep mode. */
3869 	if ((sc->rl_flags & RL_FLAG_MACSLEEP) != 0) {
3870 		if ((CSR_READ_1(sc, RL_MACDBG) & 0x80) == 0x80)
3871 			CSR_WRITE_1(sc, RL_GPIO,
3872 			    CSR_READ_1(sc, RL_GPIO) & ~0x01);
3873 	}
3874 	if ((ifp->if_capenable & IFCAP_WOL) != 0) {
3875 		if ((sc->rl_flags & RL_FLAG_8168G_PLUS) != 0) {
3876 			/* Disable RXDV gate. */
3877 			CSR_WRITE_4(sc, RL_MISC, CSR_READ_4(sc, RL_MISC) &
3878 			    ~0x00080000);
3879 		}
3880 		re_set_rxmode(sc);
3881 		if ((sc->rl_flags & RL_FLAG_WOL_MANLINK) != 0)
3882 			re_set_linkspeed(sc);
3883 		if ((sc->rl_flags & RL_FLAG_WOLRXENB) != 0)
3884 			CSR_WRITE_1(sc, RL_COMMAND, RL_CMD_RX_ENB);
3885 	}
3886 	/* Enable config register write. */
3887 	CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
3888 
3889 	/* Enable PME. */
3890 	v = CSR_READ_1(sc, sc->rl_cfg1);
3891 	v &= ~RL_CFG1_PME;
3892 	if ((ifp->if_capenable & IFCAP_WOL) != 0)
3893 		v |= RL_CFG1_PME;
3894 	CSR_WRITE_1(sc, sc->rl_cfg1, v);
3895 
3896 	v = CSR_READ_1(sc, sc->rl_cfg3);
3897 	v &= ~(RL_CFG3_WOL_LINK | RL_CFG3_WOL_MAGIC);
3898 	if ((ifp->if_capenable & IFCAP_WOL_MAGIC) != 0)
3899 		v |= RL_CFG3_WOL_MAGIC;
3900 	CSR_WRITE_1(sc, sc->rl_cfg3, v);
3901 
3902 	v = CSR_READ_1(sc, sc->rl_cfg5);
3903 	v &= ~(RL_CFG5_WOL_BCAST | RL_CFG5_WOL_MCAST | RL_CFG5_WOL_UCAST |
3904 	    RL_CFG5_WOL_LANWAKE);
3905 	if ((ifp->if_capenable & IFCAP_WOL_UCAST) != 0)
3906 		v |= RL_CFG5_WOL_UCAST;
3907 	if ((ifp->if_capenable & IFCAP_WOL_MCAST) != 0)
3908 		v |= RL_CFG5_WOL_MCAST | RL_CFG5_WOL_BCAST;
3909 	if ((ifp->if_capenable & IFCAP_WOL) != 0)
3910 		v |= RL_CFG5_WOL_LANWAKE;
3911 	CSR_WRITE_1(sc, sc->rl_cfg5, v);
3912 
3913 	/* Config register write done. */
3914 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
3915 
3916 	if ((ifp->if_capenable & IFCAP_WOL) == 0 &&
3917 	    (sc->rl_flags & RL_FLAG_PHYWAKE_PM) != 0)
3918 		CSR_WRITE_1(sc, RL_PMCH, CSR_READ_1(sc, RL_PMCH) & ~0x80);
3919 	/*
3920 	 * It seems that hardware resets its link speed to 100Mbps in
3921 	 * power down mode so switching to 100Mbps in driver is not
3922 	 * needed.
3923 	 */
3924 
3925 	/* Request PME if WOL is requested. */
3926 	pmstat = pci_read_config(sc->rl_dev, pmc + PCIR_POWER_STATUS, 2);
3927 	pmstat &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3928 	if ((ifp->if_capenable & IFCAP_WOL) != 0)
3929 		pmstat |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3930 	pci_write_config(sc->rl_dev, pmc + PCIR_POWER_STATUS, pmstat, 2);
3931 }
3932 
3933 static void
3934 re_clrwol(struct rl_softc *sc)
3935 {
3936 	int			pmc;
3937 	uint8_t			v;
3938 
3939 	RL_LOCK_ASSERT(sc);
3940 
3941 	if (pci_find_cap(sc->rl_dev, PCIY_PMG, &pmc) != 0)
3942 		return;
3943 
3944 	/* Enable config register write. */
3945 	CSR_WRITE_1(sc, RL_EECMD, RL_EE_MODE);
3946 
3947 	v = CSR_READ_1(sc, sc->rl_cfg3);
3948 	v &= ~(RL_CFG3_WOL_LINK | RL_CFG3_WOL_MAGIC);
3949 	CSR_WRITE_1(sc, sc->rl_cfg3, v);
3950 
3951 	/* Config register write done. */
3952 	CSR_WRITE_1(sc, RL_EECMD, RL_EEMODE_OFF);
3953 
3954 	v = CSR_READ_1(sc, sc->rl_cfg5);
3955 	v &= ~(RL_CFG5_WOL_BCAST | RL_CFG5_WOL_MCAST | RL_CFG5_WOL_UCAST);
3956 	v &= ~RL_CFG5_WOL_LANWAKE;
3957 	CSR_WRITE_1(sc, sc->rl_cfg5, v);
3958 }
3959 
3960 static void
3961 re_add_sysctls(struct rl_softc *sc)
3962 {
3963 	struct sysctl_ctx_list	*ctx;
3964 	struct sysctl_oid_list	*children;
3965 	int			error;
3966 
3967 	ctx = device_get_sysctl_ctx(sc->rl_dev);
3968 	children = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->rl_dev));
3969 
3970 	SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "stats",
3971 	    CTLTYPE_INT | CTLFLAG_RW, sc, 0, re_sysctl_stats, "I",
3972 	    "Statistics Information");
3973 	if ((sc->rl_flags & (RL_FLAG_MSI | RL_FLAG_MSIX)) == 0)
3974 		return;
3975 
3976 	SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "int_rx_mod",
3977 	    CTLTYPE_INT | CTLFLAG_RW, &sc->rl_int_rx_mod, 0,
3978 	    sysctl_hw_re_int_mod, "I", "re RX interrupt moderation");
3979 	/* Pull in device tunables. */
3980 	sc->rl_int_rx_mod = RL_TIMER_DEFAULT;
3981 	error = resource_int_value(device_get_name(sc->rl_dev),
3982 	    device_get_unit(sc->rl_dev), "int_rx_mod", &sc->rl_int_rx_mod);
3983 	if (error == 0) {
3984 		if (sc->rl_int_rx_mod < RL_TIMER_MIN ||
3985 		    sc->rl_int_rx_mod > RL_TIMER_MAX) {
3986 			device_printf(sc->rl_dev, "int_rx_mod value out of "
3987 			    "range; using default: %d\n",
3988 			    RL_TIMER_DEFAULT);
3989 			sc->rl_int_rx_mod = RL_TIMER_DEFAULT;
3990 		}
3991 	}
3992 }
3993 
3994 static int
3995 re_sysctl_stats(SYSCTL_HANDLER_ARGS)
3996 {
3997 	struct rl_softc		*sc;
3998 	struct rl_stats		*stats;
3999 	int			error, i, result;
4000 
4001 	result = -1;
4002 	error = sysctl_handle_int(oidp, &result, 0, req);
4003 	if (error || req->newptr == NULL)
4004 		return (error);
4005 
4006 	if (result == 1) {
4007 		sc = (struct rl_softc *)arg1;
4008 		RL_LOCK(sc);
4009 		if ((sc->rl_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
4010 			RL_UNLOCK(sc);
4011 			goto done;
4012 		}
4013 		bus_dmamap_sync(sc->rl_ldata.rl_stag,
4014 		    sc->rl_ldata.rl_smap, BUS_DMASYNC_PREREAD);
4015 		CSR_WRITE_4(sc, RL_DUMPSTATS_HI,
4016 		    RL_ADDR_HI(sc->rl_ldata.rl_stats_addr));
4017 		CSR_WRITE_4(sc, RL_DUMPSTATS_LO,
4018 		    RL_ADDR_LO(sc->rl_ldata.rl_stats_addr));
4019 		CSR_WRITE_4(sc, RL_DUMPSTATS_LO,
4020 		    RL_ADDR_LO(sc->rl_ldata.rl_stats_addr |
4021 		    RL_DUMPSTATS_START));
4022 		for (i = RL_TIMEOUT; i > 0; i--) {
4023 			if ((CSR_READ_4(sc, RL_DUMPSTATS_LO) &
4024 			    RL_DUMPSTATS_START) == 0)
4025 				break;
4026 			DELAY(1000);
4027 		}
4028 		bus_dmamap_sync(sc->rl_ldata.rl_stag,
4029 		    sc->rl_ldata.rl_smap, BUS_DMASYNC_POSTREAD);
4030 		RL_UNLOCK(sc);
4031 		if (i == 0) {
4032 			device_printf(sc->rl_dev,
4033 			    "DUMP statistics request timed out\n");
4034 			return (ETIMEDOUT);
4035 		}
4036 done:
4037 		stats = sc->rl_ldata.rl_stats;
4038 		printf("%s statistics:\n", device_get_nameunit(sc->rl_dev));
4039 		printf("Tx frames : %ju\n",
4040 		    (uintmax_t)le64toh(stats->rl_tx_pkts));
4041 		printf("Rx frames : %ju\n",
4042 		    (uintmax_t)le64toh(stats->rl_rx_pkts));
4043 		printf("Tx errors : %ju\n",
4044 		    (uintmax_t)le64toh(stats->rl_tx_errs));
4045 		printf("Rx errors : %u\n",
4046 		    le32toh(stats->rl_rx_errs));
4047 		printf("Rx missed frames : %u\n",
4048 		    (uint32_t)le16toh(stats->rl_missed_pkts));
4049 		printf("Rx frame alignment errs : %u\n",
4050 		    (uint32_t)le16toh(stats->rl_rx_framealign_errs));
4051 		printf("Tx single collisions : %u\n",
4052 		    le32toh(stats->rl_tx_onecoll));
4053 		printf("Tx multiple collisions : %u\n",
4054 		    le32toh(stats->rl_tx_multicolls));
4055 		printf("Rx unicast frames : %ju\n",
4056 		    (uintmax_t)le64toh(stats->rl_rx_ucasts));
4057 		printf("Rx broadcast frames : %ju\n",
4058 		    (uintmax_t)le64toh(stats->rl_rx_bcasts));
4059 		printf("Rx multicast frames : %u\n",
4060 		    le32toh(stats->rl_rx_mcasts));
4061 		printf("Tx aborts : %u\n",
4062 		    (uint32_t)le16toh(stats->rl_tx_aborts));
4063 		printf("Tx underruns : %u\n",
4064 		    (uint32_t)le16toh(stats->rl_rx_underruns));
4065 	}
4066 
4067 	return (error);
4068 }
4069 
4070 static int
4071 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
4072 {
4073 	int error, value;
4074 
4075 	if (arg1 == NULL)
4076 		return (EINVAL);
4077 	value = *(int *)arg1;
4078 	error = sysctl_handle_int(oidp, &value, 0, req);
4079 	if (error || req->newptr == NULL)
4080 		return (error);
4081 	if (value < low || value > high)
4082 		return (EINVAL);
4083 	*(int *)arg1 = value;
4084 
4085 	return (0);
4086 }
4087 
4088 static int
4089 sysctl_hw_re_int_mod(SYSCTL_HANDLER_ARGS)
4090 {
4091 
4092 	return (sysctl_int_range(oidp, arg1, arg2, req, RL_TIMER_MIN,
4093 	    RL_TIMER_MAX));
4094 }
4095 
4096 #ifdef NETDUMP
4097 static void
4098 re_netdump_init(struct ifnet *ifp, int *nrxr, int *ncl, int *clsize)
4099 {
4100 	struct rl_softc *sc;
4101 
4102 	sc = if_getsoftc(ifp);
4103 	RL_LOCK(sc);
4104 	*nrxr = sc->rl_ldata.rl_rx_desc_cnt;
4105 	*ncl = NETDUMP_MAX_IN_FLIGHT;
4106 	*clsize = (ifp->if_mtu > RL_MTU &&
4107 	    (sc->rl_flags & RL_FLAG_JUMBOV2) != 0) ? MJUM9BYTES : MCLBYTES;
4108 	RL_UNLOCK(sc);
4109 }
4110 
4111 static void
4112 re_netdump_event(struct ifnet *ifp __unused, enum netdump_ev event __unused)
4113 {
4114 }
4115 
4116 static int
4117 re_netdump_transmit(struct ifnet *ifp, struct mbuf *m)
4118 {
4119 	struct rl_softc *sc;
4120 	int error;
4121 
4122 	sc = if_getsoftc(ifp);
4123 	if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
4124 	    IFF_DRV_RUNNING || (sc->rl_flags & RL_FLAG_LINK) == 0)
4125 		return (EBUSY);
4126 
4127 	error = re_encap(sc, &m);
4128 	if (error == 0)
4129 		re_start_tx(sc);
4130 	return (error);
4131 }
4132 
4133 static int
4134 re_netdump_poll(struct ifnet *ifp, int count)
4135 {
4136 	struct rl_softc *sc;
4137 	int error;
4138 
4139 	sc = if_getsoftc(ifp);
4140 	if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0 ||
4141 	    (sc->rl_flags & RL_FLAG_LINK) == 0)
4142 		return (EBUSY);
4143 
4144 	re_txeof(sc);
4145 	error = re_rxeof(sc, NULL);
4146 	if (error != 0 && error != EAGAIN)
4147 		return (error);
4148 	return (0);
4149 }
4150 #endif /* NETDUMP */
4151