Home | History | Annotate | Line # | Download | only in ic
rtl8169.c revision 1.72.2.7.2.2
      1 /*	$NetBSD: rtl8169.c,v 1.72.2.7.2.2 2008/06/03 20:47:19 skrll Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1997, 1998-2003
      5  *	Bill Paul <wpaul (at) 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 /* $FreeBSD: /repoman/r/ncvs/src/sys/dev/re/if_re.c,v 1.20 2004/04/11 20:34:08 ru Exp $ */
     37 
     38 /*
     39  * RealTek 8139C+/8169/8169S/8110S PCI NIC driver
     40  *
     41  * Written by Bill Paul <wpaul (at) 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  * four devices in this family: the RTL8139C+, the RTL8169, the RTL8169S
     50  * and the RTL8110S.
     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 7.5K, so the max MTU possible with this
    110  * driver is 7500 bytes.
    111  */
    112 
    113 #include "bpfilter.h"
    114 #include "vlan.h"
    115 
    116 #include <sys/param.h>
    117 #include <sys/endian.h>
    118 #include <sys/systm.h>
    119 #include <sys/sockio.h>
    120 #include <sys/mbuf.h>
    121 #include <sys/malloc.h>
    122 #include <sys/kernel.h>
    123 #include <sys/socket.h>
    124 #include <sys/device.h>
    125 
    126 #include <net/if.h>
    127 #include <net/if_arp.h>
    128 #include <net/if_dl.h>
    129 #include <net/if_ether.h>
    130 #include <net/if_media.h>
    131 #include <net/if_vlanvar.h>
    132 
    133 #include <netinet/in_systm.h>	/* XXX for IP_MAXPACKET */
    134 #include <netinet/in.h>		/* XXX for IP_MAXPACKET */
    135 #include <netinet/ip.h>		/* XXX for IP_MAXPACKET */
    136 
    137 #if NBPFILTER > 0
    138 #include <net/bpf.h>
    139 #endif
    140 
    141 #include <machine/bus.h>
    142 
    143 #include <dev/mii/mii.h>
    144 #include <dev/mii/miivar.h>
    145 
    146 #include <dev/pci/pcireg.h>
    147 #include <dev/pci/pcivar.h>
    148 #include <dev/pci/pcidevs.h>
    149 
    150 #include <dev/ic/rtl81x9reg.h>
    151 #include <dev/ic/rtl81x9var.h>
    152 
    153 #include <dev/ic/rtl8169var.h>
    154 
    155 static inline void re_set_bufaddr(struct re_desc *, bus_addr_t);
    156 
    157 static int re_newbuf(struct rtk_softc *, int, struct mbuf *);
    158 static int re_rx_list_init(struct rtk_softc *);
    159 static int re_tx_list_init(struct rtk_softc *);
    160 static void re_rxeof(struct rtk_softc *);
    161 static void re_txeof(struct rtk_softc *);
    162 static void re_tick(void *);
    163 static void re_start(struct ifnet *);
    164 static int re_ioctl(struct ifnet *, u_long, caddr_t);
    165 static int re_init(struct ifnet *);
    166 static void re_stop(struct ifnet *, int);
    167 static void re_watchdog(struct ifnet *);
    168 
    169 static void re_shutdown(void *);
    170 static int re_enable(struct rtk_softc *);
    171 static void re_disable(struct rtk_softc *);
    172 static void re_power(int, void *);
    173 
    174 static int re_ifmedia_upd(struct ifnet *);
    175 static void re_ifmedia_sts(struct ifnet *, struct ifmediareq *);
    176 
    177 static int re_gmii_readreg(struct device *, int, int);
    178 static void re_gmii_writereg(struct device *, int, int, int);
    179 
    180 static int re_miibus_readreg(struct device *, int, int);
    181 static void re_miibus_writereg(struct device *, int, int, int);
    182 static void re_miibus_statchg(struct device *);
    183 
    184 static void re_reset(struct rtk_softc *);
    185 
    186 static inline void
    187 re_set_bufaddr(struct re_desc *d, bus_addr_t addr)
    188 {
    189 
    190 	d->re_bufaddr_lo = htole32((uint32_t)addr);
    191 	if (sizeof(bus_addr_t) == sizeof(uint64_t))
    192 		d->re_bufaddr_hi = htole32((uint64_t)addr >> 32);
    193 	else
    194 		d->re_bufaddr_hi = 0;
    195 }
    196 
    197 static int
    198 re_gmii_readreg(struct device *self, int phy, int reg)
    199 {
    200 	struct rtk_softc	*sc = (void *)self;
    201 	uint32_t		rval;
    202 	int			i;
    203 
    204 	if (phy != 7)
    205 		return 0;
    206 
    207 	/* Let the rgephy driver read the GMEDIASTAT register */
    208 
    209 	if (reg == RTK_GMEDIASTAT) {
    210 		rval = CSR_READ_1(sc, RTK_GMEDIASTAT);
    211 		return rval;
    212 	}
    213 
    214 	CSR_WRITE_4(sc, RTK_PHYAR, reg << 16);
    215 	DELAY(1000);
    216 
    217 	for (i = 0; i < RTK_TIMEOUT; i++) {
    218 		rval = CSR_READ_4(sc, RTK_PHYAR);
    219 		if (rval & RTK_PHYAR_BUSY)
    220 			break;
    221 		DELAY(100);
    222 	}
    223 
    224 	if (i == RTK_TIMEOUT) {
    225 		aprint_error("%s: PHY read failed\n", sc->sc_dev.dv_xname);
    226 		return 0;
    227 	}
    228 
    229 	return rval & RTK_PHYAR_PHYDATA;
    230 }
    231 
    232 static void
    233 re_gmii_writereg(struct device *dev, int phy, int reg, int data)
    234 {
    235 	struct rtk_softc	*sc = (void *)dev;
    236 	uint32_t		rval;
    237 	int			i;
    238 
    239 	CSR_WRITE_4(sc, RTK_PHYAR, (reg << 16) |
    240 	    (data & RTK_PHYAR_PHYDATA) | RTK_PHYAR_BUSY);
    241 	DELAY(1000);
    242 
    243 	for (i = 0; i < RTK_TIMEOUT; i++) {
    244 		rval = CSR_READ_4(sc, RTK_PHYAR);
    245 		if (!(rval & RTK_PHYAR_BUSY))
    246 			break;
    247 		DELAY(100);
    248 	}
    249 
    250 	if (i == RTK_TIMEOUT) {
    251 		aprint_error("%s: PHY write reg %x <- %x failed\n",
    252 		    sc->sc_dev.dv_xname, reg, data);
    253 	}
    254 }
    255 
    256 static int
    257 re_miibus_readreg(struct device *dev, int phy, int reg)
    258 {
    259 	struct rtk_softc	*sc = (void *)dev;
    260 	uint16_t		rval = 0;
    261 	uint16_t		re8139_reg = 0;
    262 	int			s;
    263 
    264 	s = splnet();
    265 
    266 	if ((sc->sc_quirk & RTKQ_8139CPLUS) == 0) {
    267 		rval = re_gmii_readreg(dev, phy, reg);
    268 		splx(s);
    269 		return rval;
    270 	}
    271 
    272 	/* Pretend the internal PHY is only at address 0 */
    273 	if (phy) {
    274 		splx(s);
    275 		return 0;
    276 	}
    277 	switch (reg) {
    278 	case MII_BMCR:
    279 		re8139_reg = RTK_BMCR;
    280 		break;
    281 	case MII_BMSR:
    282 		re8139_reg = RTK_BMSR;
    283 		break;
    284 	case MII_ANAR:
    285 		re8139_reg = RTK_ANAR;
    286 		break;
    287 	case MII_ANER:
    288 		re8139_reg = RTK_ANER;
    289 		break;
    290 	case MII_ANLPAR:
    291 		re8139_reg = RTK_LPAR;
    292 		break;
    293 	case MII_PHYIDR1:
    294 	case MII_PHYIDR2:
    295 		splx(s);
    296 		return 0;
    297 	/*
    298 	 * Allow the rlphy driver to read the media status
    299 	 * register. If we have a link partner which does not
    300 	 * support NWAY, this is the register which will tell
    301 	 * us the results of parallel detection.
    302 	 */
    303 	case RTK_MEDIASTAT:
    304 		rval = CSR_READ_1(sc, RTK_MEDIASTAT);
    305 		splx(s);
    306 		return rval;
    307 	default:
    308 		aprint_error("%s: bad phy register\n", sc->sc_dev.dv_xname);
    309 		splx(s);
    310 		return 0;
    311 	}
    312 	rval = CSR_READ_2(sc, re8139_reg);
    313 	if ((sc->sc_quirk & RTKQ_8139CPLUS) != 0 && re8139_reg == RTK_BMCR) {
    314 		/* 8139C+ has different bit layout. */
    315 		rval &= ~(BMCR_LOOP | BMCR_ISO);
    316 	}
    317 	splx(s);
    318 	return rval;
    319 }
    320 
    321 static void
    322 re_miibus_writereg(struct device *dev, int phy, int reg, int data)
    323 {
    324 	struct rtk_softc	*sc = (void *)dev;
    325 	uint16_t		re8139_reg = 0;
    326 	int			s;
    327 
    328 	s = splnet();
    329 
    330 	if ((sc->sc_quirk & RTKQ_8139CPLUS) == 0) {
    331 		re_gmii_writereg(dev, phy, reg, data);
    332 		splx(s);
    333 		return;
    334 	}
    335 
    336 	/* Pretend the internal PHY is only at address 0 */
    337 	if (phy) {
    338 		splx(s);
    339 		return;
    340 	}
    341 	switch (reg) {
    342 	case MII_BMCR:
    343 		re8139_reg = RTK_BMCR;
    344 		if ((sc->sc_quirk & RTKQ_8139CPLUS) != 0) {
    345 			/* 8139C+ has different bit layout. */
    346 			data &= ~(BMCR_LOOP | BMCR_ISO);
    347 		}
    348 		break;
    349 	case MII_BMSR:
    350 		re8139_reg = RTK_BMSR;
    351 		break;
    352 	case MII_ANAR:
    353 		re8139_reg = RTK_ANAR;
    354 		break;
    355 	case MII_ANER:
    356 		re8139_reg = RTK_ANER;
    357 		break;
    358 	case MII_ANLPAR:
    359 		re8139_reg = RTK_LPAR;
    360 		break;
    361 	case MII_PHYIDR1:
    362 	case MII_PHYIDR2:
    363 		splx(s);
    364 		return;
    365 		break;
    366 	default:
    367 		aprint_error("%s: bad phy register\n", sc->sc_dev.dv_xname);
    368 		splx(s);
    369 		return;
    370 	}
    371 	CSR_WRITE_2(sc, re8139_reg, data);
    372 	splx(s);
    373 	return;
    374 }
    375 
    376 static void
    377 re_miibus_statchg(struct device *dev)
    378 {
    379 
    380 	return;
    381 }
    382 
    383 static void
    384 re_reset(struct rtk_softc *sc)
    385 {
    386 	int		i;
    387 
    388 	CSR_WRITE_1(sc, RTK_COMMAND, RTK_CMD_RESET);
    389 
    390 	for (i = 0; i < RTK_TIMEOUT; i++) {
    391 		DELAY(10);
    392 		if ((CSR_READ_1(sc, RTK_COMMAND) & RTK_CMD_RESET) == 0)
    393 			break;
    394 	}
    395 	if (i == RTK_TIMEOUT)
    396 		aprint_error("%s: reset never completed!\n",
    397 		    sc->sc_dev.dv_xname);
    398 
    399 	/*
    400 	 * NB: Realtek-supplied Linux driver does this only for
    401 	 * MCFG_METHOD_2, which corresponds to sc->sc_rev == 2.
    402 	 */
    403 	if (1) /* XXX check softc flag for 8169s version */
    404 		CSR_WRITE_1(sc, RTK_LDPS, 1);
    405 
    406 	return;
    407 }
    408 
    409 /*
    410  * The following routine is designed to test for a defect on some
    411  * 32-bit 8169 cards. Some of these NICs have the REQ64# and ACK64#
    412  * lines connected to the bus, however for a 32-bit only card, they
    413  * should be pulled high. The result of this defect is that the
    414  * NIC will not work right if you plug it into a 64-bit slot: DMA
    415  * operations will be done with 64-bit transfers, which will fail
    416  * because the 64-bit data lines aren't connected.
    417  *
    418  * There's no way to work around this (short of talking a soldering
    419  * iron to the board), however we can detect it. The method we use
    420  * here is to put the NIC into digital loopback mode, set the receiver
    421  * to promiscuous mode, and then try to send a frame. We then compare
    422  * the frame data we sent to what was received. If the data matches,
    423  * then the NIC is working correctly, otherwise we know the user has
    424  * a defective NIC which has been mistakenly plugged into a 64-bit PCI
    425  * slot. In the latter case, there's no way the NIC can work correctly,
    426  * so we print out a message on the console and abort the device attach.
    427  */
    428 
    429 int
    430 re_diag(struct rtk_softc *sc)
    431 {
    432 	struct ifnet		*ifp = &sc->ethercom.ec_if;
    433 	struct mbuf		*m0;
    434 	struct ether_header	*eh;
    435 	struct re_rxsoft	*rxs;
    436 	struct re_desc		*cur_rx;
    437 	bus_dmamap_t		dmamap;
    438 	uint16_t		status;
    439 	uint32_t		rxstat;
    440 	int			total_len, i, s, error = 0;
    441 	static const uint8_t	dst[] = { 0x00, 'h', 'e', 'l', 'l', 'o' };
    442 	static const uint8_t	src[] = { 0x00, 'w', 'o', 'r', 'l', 'd' };
    443 
    444 	/* Allocate a single mbuf */
    445 
    446 	MGETHDR(m0, M_DONTWAIT, MT_DATA);
    447 	if (m0 == NULL)
    448 		return ENOBUFS;
    449 
    450 	/*
    451 	 * Initialize the NIC in test mode. This sets the chip up
    452 	 * so that it can send and receive frames, but performs the
    453 	 * following special functions:
    454 	 * - Puts receiver in promiscuous mode
    455 	 * - Enables digital loopback mode
    456 	 * - Leaves interrupts turned off
    457 	 */
    458 
    459 	ifp->if_flags |= IFF_PROMISC;
    460 	sc->re_testmode = 1;
    461 	re_init(ifp);
    462 	re_stop(ifp, 0);
    463 	DELAY(100000);
    464 	re_init(ifp);
    465 
    466 	/* Put some data in the mbuf */
    467 
    468 	eh = mtod(m0, struct ether_header *);
    469 	memcpy(eh->ether_dhost, (char *)&dst, ETHER_ADDR_LEN);
    470 	memcpy(eh->ether_shost, (char *)&src, ETHER_ADDR_LEN);
    471 	eh->ether_type = htons(ETHERTYPE_IP);
    472 	m0->m_pkthdr.len = m0->m_len = ETHER_MIN_LEN - ETHER_CRC_LEN;
    473 
    474 	/*
    475 	 * Queue the packet, start transmission.
    476 	 */
    477 
    478 	CSR_WRITE_2(sc, RTK_ISR, 0xFFFF);
    479 	s = splnet();
    480 	IF_ENQUEUE(&ifp->if_snd, m0);
    481 	re_start(ifp);
    482 	splx(s);
    483 	m0 = NULL;
    484 
    485 	/* Wait for it to propagate through the chip */
    486 
    487 	DELAY(100000);
    488 	for (i = 0; i < RTK_TIMEOUT; i++) {
    489 		status = CSR_READ_2(sc, RTK_ISR);
    490 		if ((status & (RTK_ISR_TIMEOUT_EXPIRED | RTK_ISR_RX_OK)) ==
    491 		    (RTK_ISR_TIMEOUT_EXPIRED | RTK_ISR_RX_OK))
    492 			break;
    493 		DELAY(10);
    494 	}
    495 	if (i == RTK_TIMEOUT) {
    496 		aprint_error("%s: diagnostic failed, failed to receive packet "
    497 		    "in loopback mode\n", sc->sc_dev.dv_xname);
    498 		error = EIO;
    499 		goto done;
    500 	}
    501 
    502 	/*
    503 	 * The packet should have been dumped into the first
    504 	 * entry in the RX DMA ring. Grab it from there.
    505 	 */
    506 
    507 	rxs = &sc->re_ldata.re_rxsoft[0];
    508 	dmamap = rxs->rxs_dmamap;
    509 	bus_dmamap_sync(sc->sc_dmat, dmamap, 0, dmamap->dm_mapsize,
    510 	    BUS_DMASYNC_POSTREAD);
    511 	bus_dmamap_unload(sc->sc_dmat, dmamap);
    512 
    513 	m0 = rxs->rxs_mbuf;
    514 	rxs->rxs_mbuf = NULL;
    515 	eh = mtod(m0, struct ether_header *);
    516 
    517 	RE_RXDESCSYNC(sc, 0, BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
    518 	cur_rx = &sc->re_ldata.re_rx_list[0];
    519 	rxstat = le32toh(cur_rx->re_cmdstat);
    520 	total_len = rxstat & sc->re_rxlenmask;
    521 
    522 	if (total_len != ETHER_MIN_LEN) {
    523 		aprint_error("%s: diagnostic failed, received short packet\n",
    524 		    sc->sc_dev.dv_xname);
    525 		error = EIO;
    526 		goto done;
    527 	}
    528 
    529 	/* Test that the received packet data matches what we sent. */
    530 
    531 	if (memcmp((char *)&eh->ether_dhost, (char *)&dst, ETHER_ADDR_LEN) ||
    532 	    memcmp((char *)&eh->ether_shost, (char *)&src, ETHER_ADDR_LEN) ||
    533 	    ntohs(eh->ether_type) != ETHERTYPE_IP) {
    534 		aprint_error("%s: WARNING, DMA FAILURE!\n",
    535 		    sc->sc_dev.dv_xname);
    536 		aprint_error("%s: expected TX data: %s",
    537 		    sc->sc_dev.dv_xname, ether_sprintf(dst));
    538 		aprint_error("/%s/0x%x\n", ether_sprintf(src), ETHERTYPE_IP);
    539 		aprint_error("%s: received RX data: %s",
    540 		    sc->sc_dev.dv_xname,
    541 		    ether_sprintf(eh->ether_dhost));
    542 		aprint_error("/%s/0x%x\n", ether_sprintf(eh->ether_shost),
    543 		    ntohs(eh->ether_type));
    544 		aprint_error("%s: You may have a defective 32-bit NIC plugged "
    545 		    "into a 64-bit PCI slot.\n", sc->sc_dev.dv_xname);
    546 		aprint_error("%s: Please re-install the NIC in a 32-bit slot "
    547 		    "for proper operation.\n", sc->sc_dev.dv_xname);
    548 		aprint_error("%s: Read the re(4) man page for more details.\n",
    549 		    sc->sc_dev.dv_xname);
    550 		error = EIO;
    551 	}
    552 
    553  done:
    554 	/* Turn interface off, release resources */
    555 
    556 	sc->re_testmode = 0;
    557 	ifp->if_flags &= ~IFF_PROMISC;
    558 	re_stop(ifp, 0);
    559 	if (m0 != NULL)
    560 		m_freem(m0);
    561 
    562 	return error;
    563 }
    564 
    565 
    566 /*
    567  * Attach the interface. Allocate softc structures, do ifmedia
    568  * setup and ethernet/BPF attach.
    569  */
    570 void
    571 re_attach(struct rtk_softc *sc)
    572 {
    573 	u_char			eaddr[ETHER_ADDR_LEN];
    574 	uint16_t		val;
    575 	struct ifnet		*ifp;
    576 	int			error = 0, i, addr_len;
    577 
    578 	/* Reset the adapter. */
    579 	re_reset(sc);
    580 
    581 	if (rtk_read_eeprom(sc, RTK_EE_ID, RTK_EEADDR_LEN1) == 0x8129)
    582 		addr_len = RTK_EEADDR_LEN1;
    583 	else
    584 		addr_len = RTK_EEADDR_LEN0;
    585 
    586 	/*
    587 	 * Get station address from the EEPROM.
    588 	 */
    589 	for (i = 0; i < 3; i++) {
    590 		val = rtk_read_eeprom(sc, RTK_EE_EADDR0 + i, addr_len);
    591 		eaddr[(i * 2) + 0] = val & 0xff;
    592 		eaddr[(i * 2) + 1] = val >> 8;
    593 	}
    594 
    595 	if ((sc->sc_quirk & RTKQ_8139CPLUS) == 0) {
    596 		uint32_t hwrev;
    597 
    598 		/* Revision of 8169/8169S/8110s in bits 30..26, 23 */
    599 		hwrev = CSR_READ_4(sc, RTK_TXCFG) & RTK_TXCFG_HWREV;
    600 		/* These rev numbers are taken from Realtek's driver */
    601 		if (       hwrev == RTK_HWREV_8100E_SPIN2) {
    602 			sc->sc_rev = 15;
    603 		} else if (hwrev == RTK_HWREV_8100E) {
    604 			sc->sc_rev = 14;
    605 		} else if (hwrev == RTK_HWREV_8101E) {
    606 			sc->sc_rev = 13;
    607 		} else if (hwrev == RTK_HWREV_8168_SPIN2 ||
    608 		           hwrev == RTK_HWREV_8168_SPIN3) {
    609 			sc->sc_rev = 12;
    610 		} else if (hwrev == RTK_HWREV_8168_SPIN1) {
    611 			sc->sc_rev = 11;
    612 		} else if (hwrev == RTK_HWREV_8169_8110SC) {
    613 			sc->sc_rev = 5;
    614 		} else if (hwrev == RTK_HWREV_8169_8110SB) {
    615 			sc->sc_rev = 4;
    616 		} else if (hwrev == RTK_HWREV_8169S) {
    617 			sc->sc_rev = 3;
    618 		} else if (hwrev == RTK_HWREV_8110S) {
    619 			sc->sc_rev = 2;
    620 		} else if (hwrev == RTK_HWREV_8169) {
    621 			sc->sc_rev = 1;
    622 			sc->sc_quirk |= RTKQ_8169NONS;
    623 		} else {
    624 			aprint_normal("%s: Unknown revision (0x%08x)\n",
    625 			    sc->sc_dev.dv_xname, hwrev);
    626 			/* assume the latest one */
    627 			sc->sc_rev = 15;
    628 		}
    629 
    630 		/* Set RX length mask */
    631 		sc->re_rxlenmask = RE_RDESC_STAT_GFRAGLEN;
    632 		sc->re_ldata.re_tx_desc_cnt = RE_TX_DESC_CNT_8169;
    633 	} else {
    634 		/* Set RX length mask */
    635 		sc->re_rxlenmask = RE_RDESC_STAT_FRAGLEN;
    636 		sc->re_ldata.re_tx_desc_cnt = RE_TX_DESC_CNT_8139;
    637 	}
    638 
    639 	aprint_normal("%s: Ethernet address %s\n",
    640 	    sc->sc_dev.dv_xname, ether_sprintf(eaddr));
    641 
    642 	if (sc->re_ldata.re_tx_desc_cnt >
    643 	    PAGE_SIZE / sizeof(struct re_desc)) {
    644 		sc->re_ldata.re_tx_desc_cnt =
    645 		    PAGE_SIZE / sizeof(struct re_desc);
    646 	}
    647 
    648 	aprint_verbose("%s: using %d tx descriptors\n",
    649 	    sc->sc_dev.dv_xname, sc->re_ldata.re_tx_desc_cnt);
    650 	KASSERT(RE_NEXT_TX_DESC(sc, RE_TX_DESC_CNT(sc) - 1) == 0);
    651 
    652 	/* Allocate DMA'able memory for the TX ring */
    653 	if ((error = bus_dmamem_alloc(sc->sc_dmat, RE_TX_LIST_SZ(sc),
    654 	    RE_RING_ALIGN, 0, &sc->re_ldata.re_tx_listseg, 1,
    655 	    &sc->re_ldata.re_tx_listnseg, BUS_DMA_NOWAIT)) != 0) {
    656 		aprint_error("%s: can't allocate tx listseg, error = %d\n",
    657 		    sc->sc_dev.dv_xname, error);
    658 		goto fail_0;
    659 	}
    660 
    661 	/* Load the map for the TX ring. */
    662 	if ((error = bus_dmamem_map(sc->sc_dmat, &sc->re_ldata.re_tx_listseg,
    663 	    sc->re_ldata.re_tx_listnseg, RE_TX_LIST_SZ(sc),
    664 	    (caddr_t *)&sc->re_ldata.re_tx_list,
    665 	    BUS_DMA_COHERENT | BUS_DMA_NOWAIT)) != 0) {
    666 		aprint_error("%s: can't map tx list, error = %d\n",
    667 		    sc->sc_dev.dv_xname, error);
    668 	  	goto fail_1;
    669 	}
    670 	memset(sc->re_ldata.re_tx_list, 0, RE_TX_LIST_SZ(sc));
    671 
    672 	if ((error = bus_dmamap_create(sc->sc_dmat, RE_TX_LIST_SZ(sc), 1,
    673 	    RE_TX_LIST_SZ(sc), 0, 0,
    674 	    &sc->re_ldata.re_tx_list_map)) != 0) {
    675 		aprint_error("%s: can't create tx list map, error = %d\n",
    676 		    sc->sc_dev.dv_xname, error);
    677 		goto fail_2;
    678 	}
    679 
    680 
    681 	if ((error = bus_dmamap_load(sc->sc_dmat,
    682 	    sc->re_ldata.re_tx_list_map, sc->re_ldata.re_tx_list,
    683 	    RE_TX_LIST_SZ(sc), NULL, BUS_DMA_NOWAIT)) != 0) {
    684 		aprint_error("%s: can't load tx list, error = %d\n",
    685 		    sc->sc_dev.dv_xname, error);
    686 		goto fail_3;
    687 	}
    688 
    689 	/* Create DMA maps for TX buffers */
    690 	for (i = 0; i < RE_TX_QLEN; i++) {
    691 		error = bus_dmamap_create(sc->sc_dmat,
    692 		    round_page(IP_MAXPACKET),
    693 		    RE_TX_DESC_CNT(sc), RE_TDESC_CMD_FRAGLEN,
    694 		    0, 0, &sc->re_ldata.re_txq[i].txq_dmamap);
    695 		if (error) {
    696 			aprint_error("%s: can't create DMA map for TX\n",
    697 			    sc->sc_dev.dv_xname);
    698 			goto fail_4;
    699 		}
    700 	}
    701 
    702 	/* Allocate DMA'able memory for the RX ring */
    703 	/* XXX see also a comment about RE_RX_DMAMEM_SZ in rtl81x9var.h */
    704 	if ((error = bus_dmamem_alloc(sc->sc_dmat,
    705 	    RE_RX_DMAMEM_SZ, RE_RING_ALIGN, 0, &sc->re_ldata.re_rx_listseg, 1,
    706 	    &sc->re_ldata.re_rx_listnseg, BUS_DMA_NOWAIT)) != 0) {
    707 		aprint_error("%s: can't allocate rx listseg, error = %d\n",
    708 		    sc->sc_dev.dv_xname, error);
    709 		goto fail_4;
    710 	}
    711 
    712 	/* Load the map for the RX ring. */
    713 	if ((error = bus_dmamem_map(sc->sc_dmat, &sc->re_ldata.re_rx_listseg,
    714 	    sc->re_ldata.re_rx_listnseg, RE_RX_DMAMEM_SZ,
    715 	    (caddr_t *)&sc->re_ldata.re_rx_list,
    716 	    BUS_DMA_COHERENT | BUS_DMA_NOWAIT)) != 0) {
    717 		aprint_error("%s: can't map rx list, error = %d\n",
    718 		    sc->sc_dev.dv_xname, error);
    719 		goto fail_5;
    720 	}
    721 	memset(sc->re_ldata.re_rx_list, 0, RE_RX_DMAMEM_SZ);
    722 
    723 	if ((error = bus_dmamap_create(sc->sc_dmat,
    724 	    RE_RX_DMAMEM_SZ, 1, RE_RX_DMAMEM_SZ, 0, 0,
    725 	    &sc->re_ldata.re_rx_list_map)) != 0) {
    726 		aprint_error("%s: can't create rx list map, error = %d\n",
    727 		    sc->sc_dev.dv_xname, error);
    728 		goto fail_6;
    729 	}
    730 
    731 	if ((error = bus_dmamap_load(sc->sc_dmat,
    732 	    sc->re_ldata.re_rx_list_map, sc->re_ldata.re_rx_list,
    733 	    RE_RX_DMAMEM_SZ, NULL, BUS_DMA_NOWAIT)) != 0) {
    734 		aprint_error("%s: can't load rx list, error = %d\n",
    735 		    sc->sc_dev.dv_xname, error);
    736 		goto fail_7;
    737 	}
    738 
    739 	/* Create DMA maps for RX buffers */
    740 	for (i = 0; i < RE_RX_DESC_CNT; i++) {
    741 		error = bus_dmamap_create(sc->sc_dmat, MCLBYTES, 1, MCLBYTES,
    742 		    0, 0, &sc->re_ldata.re_rxsoft[i].rxs_dmamap);
    743 		if (error) {
    744 			aprint_error("%s: can't create DMA map for RX\n",
    745 			    sc->sc_dev.dv_xname);
    746 			goto fail_8;
    747 		}
    748 	}
    749 
    750 	/*
    751 	 * Record interface as attached. From here, we should not fail.
    752 	 */
    753 	sc->sc_flags |= RTK_ATTACHED;
    754 
    755 	ifp = &sc->ethercom.ec_if;
    756 	ifp->if_softc = sc;
    757 	strcpy(ifp->if_xname, sc->sc_dev.dv_xname);
    758 	ifp->if_mtu = ETHERMTU;
    759 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
    760 	ifp->if_ioctl = re_ioctl;
    761 	sc->ethercom.ec_capabilities |=
    762 	    ETHERCAP_VLAN_MTU | ETHERCAP_VLAN_HWTAGGING;
    763 	ifp->if_start = re_start;
    764 	ifp->if_stop = re_stop;
    765 
    766 	/*
    767 	 * IFCAP_CSUM_IPv4_Tx on re(4) is broken for small packets,
    768 	 * so we have a workaround to handle the bug by padding
    769 	 * such packets manually.
    770 	 */
    771 	ifp->if_capabilities |=
    772 	    IFCAP_CSUM_IPv4_Tx | IFCAP_CSUM_IPv4_Rx |
    773 	    IFCAP_CSUM_TCPv4_Tx | IFCAP_CSUM_TCPv4_Rx |
    774 	    IFCAP_CSUM_UDPv4_Tx | IFCAP_CSUM_UDPv4_Rx |
    775 	    IFCAP_TSOv4;
    776 	ifp->if_watchdog = re_watchdog;
    777 	ifp->if_init = re_init;
    778 	ifp->if_snd.ifq_maxlen = RE_IFQ_MAXLEN;
    779 	ifp->if_capenable = ifp->if_capabilities;
    780 	IFQ_SET_READY(&ifp->if_snd);
    781 
    782 	callout_init(&sc->rtk_tick_ch);
    783 
    784 	/* Do MII setup */
    785 	sc->mii.mii_ifp = ifp;
    786 	sc->mii.mii_readreg = re_miibus_readreg;
    787 	sc->mii.mii_writereg = re_miibus_writereg;
    788 	sc->mii.mii_statchg = re_miibus_statchg;
    789 	ifmedia_init(&sc->mii.mii_media, IFM_IMASK, re_ifmedia_upd,
    790 	    re_ifmedia_sts);
    791 	mii_attach(&sc->sc_dev, &sc->mii, 0xffffffff, MII_PHY_ANY,
    792 	    MII_OFFSET_ANY, 0);
    793 	ifmedia_set(&sc->mii.mii_media, IFM_ETHER | IFM_AUTO);
    794 
    795 	/*
    796 	 * Call MI attach routine.
    797 	 */
    798 	if_attach(ifp);
    799 	ether_ifattach(ifp, eaddr);
    800 
    801 
    802 	/*
    803 	 * Make sure the interface is shutdown during reboot.
    804 	 */
    805 	sc->sc_sdhook = shutdownhook_establish(re_shutdown, sc);
    806 	if (sc->sc_sdhook == NULL)
    807 		aprint_error("%s: WARNING: unable to establish shutdown hook\n",
    808 		    sc->sc_dev.dv_xname);
    809 	/*
    810 	 * Add a suspend hook to make sure we come back up after a
    811 	 * resume.
    812 	 */
    813 	sc->sc_powerhook = powerhook_establish(sc->sc_dev.dv_xname,
    814 	    re_power, sc);
    815 	if (sc->sc_powerhook == NULL)
    816 		aprint_error("%s: WARNING: unable to establish power hook\n",
    817 		    sc->sc_dev.dv_xname);
    818 
    819 
    820 	return;
    821 
    822  fail_8:
    823 	/* Destroy DMA maps for RX buffers. */
    824 	for (i = 0; i < RE_RX_DESC_CNT; i++)
    825 		if (sc->re_ldata.re_rxsoft[i].rxs_dmamap != NULL)
    826 			bus_dmamap_destroy(sc->sc_dmat,
    827 			    sc->re_ldata.re_rxsoft[i].rxs_dmamap);
    828 
    829 	/* Free DMA'able memory for the RX ring. */
    830 	bus_dmamap_unload(sc->sc_dmat, sc->re_ldata.re_rx_list_map);
    831  fail_7:
    832 	bus_dmamap_destroy(sc->sc_dmat, sc->re_ldata.re_rx_list_map);
    833  fail_6:
    834 	bus_dmamem_unmap(sc->sc_dmat,
    835 	    (caddr_t)sc->re_ldata.re_rx_list, RE_RX_DMAMEM_SZ);
    836  fail_5:
    837 	bus_dmamem_free(sc->sc_dmat,
    838 	    &sc->re_ldata.re_rx_listseg, sc->re_ldata.re_rx_listnseg);
    839 
    840  fail_4:
    841 	/* Destroy DMA maps for TX buffers. */
    842 	for (i = 0; i < RE_TX_QLEN; i++)
    843 		if (sc->re_ldata.re_txq[i].txq_dmamap != NULL)
    844 			bus_dmamap_destroy(sc->sc_dmat,
    845 			    sc->re_ldata.re_txq[i].txq_dmamap);
    846 
    847 	/* Free DMA'able memory for the TX ring. */
    848 	bus_dmamap_unload(sc->sc_dmat, sc->re_ldata.re_tx_list_map);
    849  fail_3:
    850 	bus_dmamap_destroy(sc->sc_dmat, sc->re_ldata.re_tx_list_map);
    851  fail_2:
    852 	bus_dmamem_unmap(sc->sc_dmat,
    853 	    (caddr_t)sc->re_ldata.re_tx_list, RE_TX_LIST_SZ(sc));
    854  fail_1:
    855 	bus_dmamem_free(sc->sc_dmat,
    856 	    &sc->re_ldata.re_tx_listseg, sc->re_ldata.re_tx_listnseg);
    857  fail_0:
    858 	return;
    859 }
    860 
    861 
    862 /*
    863  * re_activate:
    864  *     Handle device activation/deactivation requests.
    865  */
    866 int
    867 re_activate(struct device *self, enum devact act)
    868 {
    869 	struct rtk_softc *sc = (void *)self;
    870 	int s, error = 0;
    871 
    872 	s = splnet();
    873 	switch (act) {
    874 	case DVACT_ACTIVATE:
    875 		error = EOPNOTSUPP;
    876 		break;
    877 	case DVACT_DEACTIVATE:
    878 		mii_activate(&sc->mii, act, MII_PHY_ANY, MII_OFFSET_ANY);
    879 		if_deactivate(&sc->ethercom.ec_if);
    880 		break;
    881 	}
    882 	splx(s);
    883 
    884 	return error;
    885 }
    886 
    887 /*
    888  * re_detach:
    889  *     Detach a rtk interface.
    890  */
    891 int
    892 re_detach(struct rtk_softc *sc)
    893 {
    894 	struct ifnet *ifp = &sc->ethercom.ec_if;
    895 	int i;
    896 
    897 	/*
    898 	 * Succeed now if there isn't any work to do.
    899 	 */
    900 	if ((sc->sc_flags & RTK_ATTACHED) == 0)
    901 		return 0;
    902 
    903 	/* Unhook our tick handler. */
    904 	callout_stop(&sc->rtk_tick_ch);
    905 
    906 	/* Detach all PHYs. */
    907 	mii_detach(&sc->mii, MII_PHY_ANY, MII_OFFSET_ANY);
    908 
    909 	/* Delete all remaining media. */
    910 	ifmedia_delete_instance(&sc->mii.mii_media, IFM_INST_ANY);
    911 
    912 	ether_ifdetach(ifp);
    913 	if_detach(ifp);
    914 
    915 	/* Destroy DMA maps for RX buffers. */
    916 	for (i = 0; i < RE_RX_DESC_CNT; i++)
    917 		if (sc->re_ldata.re_rxsoft[i].rxs_dmamap != NULL)
    918 			bus_dmamap_destroy(sc->sc_dmat,
    919 			    sc->re_ldata.re_rxsoft[i].rxs_dmamap);
    920 
    921 	/* Free DMA'able memory for the RX ring. */
    922 	bus_dmamap_unload(sc->sc_dmat, sc->re_ldata.re_rx_list_map);
    923 	bus_dmamap_destroy(sc->sc_dmat, sc->re_ldata.re_rx_list_map);
    924 	bus_dmamem_unmap(sc->sc_dmat,
    925 	    (caddr_t)sc->re_ldata.re_rx_list, RE_RX_DMAMEM_SZ);
    926 	bus_dmamem_free(sc->sc_dmat,
    927 	    &sc->re_ldata.re_rx_listseg, sc->re_ldata.re_rx_listnseg);
    928 
    929 	/* Destroy DMA maps for TX buffers. */
    930 	for (i = 0; i < RE_TX_QLEN; i++)
    931 		if (sc->re_ldata.re_txq[i].txq_dmamap != NULL)
    932 			bus_dmamap_destroy(sc->sc_dmat,
    933 			    sc->re_ldata.re_txq[i].txq_dmamap);
    934 
    935 	/* Free DMA'able memory for the TX ring. */
    936 	bus_dmamap_unload(sc->sc_dmat, sc->re_ldata.re_tx_list_map);
    937 	bus_dmamap_destroy(sc->sc_dmat, sc->re_ldata.re_tx_list_map);
    938 	bus_dmamem_unmap(sc->sc_dmat,
    939 	    (caddr_t)sc->re_ldata.re_tx_list, RE_TX_LIST_SZ(sc));
    940 	bus_dmamem_free(sc->sc_dmat,
    941 	    &sc->re_ldata.re_tx_listseg, sc->re_ldata.re_tx_listnseg);
    942 
    943 
    944 	shutdownhook_disestablish(sc->sc_sdhook);
    945 	powerhook_disestablish(sc->sc_powerhook);
    946 
    947 	return 0;
    948 }
    949 
    950 /*
    951  * re_enable:
    952  *     Enable the RTL81X9 chip.
    953  */
    954 static int
    955 re_enable(struct rtk_softc *sc)
    956 {
    957 
    958 	if (RTK_IS_ENABLED(sc) == 0 && sc->sc_enable != NULL) {
    959 		if ((*sc->sc_enable)(sc) != 0) {
    960 			aprint_error("%s: device enable failed\n",
    961 			    sc->sc_dev.dv_xname);
    962 			return EIO;
    963 		}
    964 		sc->sc_flags |= RTK_ENABLED;
    965 	}
    966 	return 0;
    967 }
    968 
    969 /*
    970  * re_disable:
    971  *     Disable the RTL81X9 chip.
    972  */
    973 static void
    974 re_disable(struct rtk_softc *sc)
    975 {
    976 
    977 	if (RTK_IS_ENABLED(sc) && sc->sc_disable != NULL) {
    978 		(*sc->sc_disable)(sc);
    979 		sc->sc_flags &= ~RTK_ENABLED;
    980 	}
    981 }
    982 
    983 /*
    984  * re_power:
    985  *     Power management (suspend/resume) hook.
    986  */
    987 void
    988 re_power(int why, void *arg)
    989 {
    990 	struct rtk_softc *sc = (void *)arg;
    991 	struct ifnet *ifp = &sc->ethercom.ec_if;
    992 	int s;
    993 
    994 	s = splnet();
    995 	switch (why) {
    996 	case PWR_SUSPEND:
    997 	case PWR_STANDBY:
    998 		re_stop(ifp, 0);
    999 		if (sc->sc_power != NULL)
   1000 			(*sc->sc_power)(sc, why);
   1001 		break;
   1002 	case PWR_RESUME:
   1003 		if (ifp->if_flags & IFF_UP) {
   1004 			if (sc->sc_power != NULL)
   1005 				(*sc->sc_power)(sc, why);
   1006 			re_init(ifp);
   1007 		}
   1008 		break;
   1009 	case PWR_SOFTSUSPEND:
   1010 	case PWR_SOFTSTANDBY:
   1011 	case PWR_SOFTRESUME:
   1012 		break;
   1013 	}
   1014 	splx(s);
   1015 }
   1016 
   1017 
   1018 static int
   1019 re_newbuf(struct rtk_softc *sc, int idx, struct mbuf *m)
   1020 {
   1021 	struct mbuf		*n = NULL;
   1022 	bus_dmamap_t		map;
   1023 	struct re_desc		*d;
   1024 	struct re_rxsoft	*rxs;
   1025 	uint32_t		cmdstat;
   1026 	int			error;
   1027 
   1028 	if (m == NULL) {
   1029 		MGETHDR(n, M_DONTWAIT, MT_DATA);
   1030 		if (n == NULL)
   1031 			return ENOBUFS;
   1032 
   1033 		MCLGET(n, M_DONTWAIT);
   1034 		if ((n->m_flags & M_EXT) == 0) {
   1035 			m_freem(n);
   1036 			return ENOBUFS;
   1037 		}
   1038 		m = n;
   1039 	} else
   1040 		m->m_data = m->m_ext.ext_buf;
   1041 
   1042 	/*
   1043 	 * Initialize mbuf length fields and fixup
   1044 	 * alignment so that the frame payload is
   1045 	 * longword aligned.
   1046 	 */
   1047 	m->m_len = m->m_pkthdr.len = MCLBYTES - RE_ETHER_ALIGN;
   1048 	m->m_data += RE_ETHER_ALIGN;
   1049 
   1050 	rxs = &sc->re_ldata.re_rxsoft[idx];
   1051 	map = rxs->rxs_dmamap;
   1052 	error = bus_dmamap_load_mbuf(sc->sc_dmat, map, m,
   1053 	    BUS_DMA_READ|BUS_DMA_NOWAIT);
   1054 
   1055 	if (error)
   1056 		goto out;
   1057 
   1058 	bus_dmamap_sync(sc->sc_dmat, map, 0, map->dm_mapsize,
   1059 	    BUS_DMASYNC_PREREAD);
   1060 
   1061 	d = &sc->re_ldata.re_rx_list[idx];
   1062 #ifdef DIAGNOSTIC
   1063 	RE_RXDESCSYNC(sc, idx, BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
   1064 	cmdstat = le32toh(d->re_cmdstat);
   1065 	RE_RXDESCSYNC(sc, idx, BUS_DMASYNC_PREREAD);
   1066 	if (cmdstat & RE_RDESC_STAT_OWN) {
   1067 		panic("%s: tried to map busy RX descriptor",
   1068 		    sc->sc_dev.dv_xname);
   1069 	}
   1070 #endif
   1071 
   1072 	rxs->rxs_mbuf = m;
   1073 
   1074 	d->re_vlanctl = 0;
   1075 	cmdstat = map->dm_segs[0].ds_len;
   1076 	if (idx == (RE_RX_DESC_CNT - 1))
   1077 		cmdstat |= RE_RDESC_CMD_EOR;
   1078 	re_set_bufaddr(d, map->dm_segs[0].ds_addr);
   1079 	d->re_cmdstat = htole32(cmdstat);
   1080 	RE_RXDESCSYNC(sc, idx, BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
   1081 	cmdstat |= RE_RDESC_CMD_OWN;
   1082 	d->re_cmdstat = htole32(cmdstat);
   1083 	RE_RXDESCSYNC(sc, idx, BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
   1084 
   1085 	return 0;
   1086  out:
   1087 	if (n != NULL)
   1088 		m_freem(n);
   1089 	return ENOMEM;
   1090 }
   1091 
   1092 static int
   1093 re_tx_list_init(struct rtk_softc *sc)
   1094 {
   1095 	int i;
   1096 
   1097 	memset(sc->re_ldata.re_tx_list, 0, RE_TX_LIST_SZ(sc));
   1098 	for (i = 0; i < RE_TX_QLEN; i++) {
   1099 		sc->re_ldata.re_txq[i].txq_mbuf = NULL;
   1100 	}
   1101 
   1102 	bus_dmamap_sync(sc->sc_dmat,
   1103 	    sc->re_ldata.re_tx_list_map, 0,
   1104 	    sc->re_ldata.re_tx_list_map->dm_mapsize,
   1105 	    BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
   1106 	sc->re_ldata.re_txq_prodidx = 0;
   1107 	sc->re_ldata.re_txq_considx = 0;
   1108 	sc->re_ldata.re_txq_free = RE_TX_QLEN;
   1109 	sc->re_ldata.re_tx_free = RE_TX_DESC_CNT(sc);
   1110 	sc->re_ldata.re_tx_nextfree = 0;
   1111 
   1112 	return 0;
   1113 }
   1114 
   1115 static int
   1116 re_rx_list_init(struct rtk_softc *sc)
   1117 {
   1118 	int			i;
   1119 
   1120 	memset((char *)sc->re_ldata.re_rx_list, 0, RE_RX_LIST_SZ);
   1121 
   1122 	for (i = 0; i < RE_RX_DESC_CNT; i++) {
   1123 		if (re_newbuf(sc, i, NULL) == ENOBUFS)
   1124 			return ENOBUFS;
   1125 	}
   1126 
   1127 	sc->re_ldata.re_rx_prodidx = 0;
   1128 	sc->re_head = sc->re_tail = NULL;
   1129 
   1130 	return 0;
   1131 }
   1132 
   1133 /*
   1134  * RX handler for C+ and 8169. For the gigE chips, we support
   1135  * the reception of jumbo frames that have been fragmented
   1136  * across multiple 2K mbuf cluster buffers.
   1137  */
   1138 static void
   1139 re_rxeof(struct rtk_softc *sc)
   1140 {
   1141 	struct mbuf		*m;
   1142 	struct ifnet		*ifp;
   1143 	int			i, total_len;
   1144 	struct re_desc		*cur_rx;
   1145 	struct re_rxsoft	*rxs;
   1146 	uint32_t		rxstat, rxvlan;
   1147 
   1148 	ifp = &sc->ethercom.ec_if;
   1149 
   1150 	for (i = sc->re_ldata.re_rx_prodidx;; i = RE_NEXT_RX_DESC(sc, i)) {
   1151 		cur_rx = &sc->re_ldata.re_rx_list[i];
   1152 		RE_RXDESCSYNC(sc, i,
   1153 		    BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
   1154 		rxstat = le32toh(cur_rx->re_cmdstat);
   1155 		rxvlan = le32toh(cur_rx->re_vlanctl);
   1156 		RE_RXDESCSYNC(sc, i, BUS_DMASYNC_PREREAD);
   1157 		if ((rxstat & RE_RDESC_STAT_OWN) != 0) {
   1158 			break;
   1159 		}
   1160 		total_len = rxstat & sc->re_rxlenmask;
   1161 		rxs = &sc->re_ldata.re_rxsoft[i];
   1162 		m = rxs->rxs_mbuf;
   1163 
   1164 		/* Invalidate the RX mbuf and unload its map */
   1165 
   1166 		bus_dmamap_sync(sc->sc_dmat,
   1167 		    rxs->rxs_dmamap, 0, rxs->rxs_dmamap->dm_mapsize,
   1168 		    BUS_DMASYNC_POSTREAD);
   1169 		bus_dmamap_unload(sc->sc_dmat, rxs->rxs_dmamap);
   1170 
   1171 		if ((rxstat & RE_RDESC_STAT_EOF) == 0) {
   1172 			m->m_len = MCLBYTES - RE_ETHER_ALIGN;
   1173 			if (sc->re_head == NULL)
   1174 				sc->re_head = sc->re_tail = m;
   1175 			else {
   1176 				m->m_flags &= ~M_PKTHDR;
   1177 				sc->re_tail->m_next = m;
   1178 				sc->re_tail = m;
   1179 			}
   1180 			re_newbuf(sc, i, NULL);
   1181 			continue;
   1182 		}
   1183 
   1184 		/*
   1185 		 * NOTE: for the 8139C+, the frame length field
   1186 		 * is always 12 bits in size, but for the gigE chips,
   1187 		 * it is 13 bits (since the max RX frame length is 16K).
   1188 		 * Unfortunately, all 32 bits in the status word
   1189 		 * were already used, so to make room for the extra
   1190 		 * length bit, RealTek took out the 'frame alignment
   1191 		 * error' bit and shifted the other status bits
   1192 		 * over one slot. The OWN, EOR, FS and LS bits are
   1193 		 * still in the same places. We have already extracted
   1194 		 * the frame length and checked the OWN bit, so rather
   1195 		 * than using an alternate bit mapping, we shift the
   1196 		 * status bits one space to the right so we can evaluate
   1197 		 * them using the 8169 status as though it was in the
   1198 		 * same format as that of the 8139C+.
   1199 		 */
   1200 		if ((sc->sc_quirk & RTKQ_8139CPLUS) == 0)
   1201 			rxstat >>= 1;
   1202 
   1203 		if (__predict_false((rxstat & RE_RDESC_STAT_RXERRSUM) != 0)) {
   1204 #ifdef RE_DEBUG
   1205 			aprint_error("%s: RX error (rxstat = 0x%08x)",
   1206 			    sc->sc_dev.dv_xname, rxstat);
   1207 			if (rxstat & RE_RDESC_STAT_FRALIGN)
   1208 				aprint_error(", frame alignment error");
   1209 			if (rxstat & RE_RDESC_STAT_BUFOFLOW)
   1210 				aprint_error(", out of buffer space");
   1211 			if (rxstat & RE_RDESC_STAT_FIFOOFLOW)
   1212 				aprint_error(", FIFO overrun");
   1213 			if (rxstat & RE_RDESC_STAT_GIANT)
   1214 				aprint_error(", giant packet");
   1215 			if (rxstat & RE_RDESC_STAT_RUNT)
   1216 				aprint_error(", runt packet");
   1217 			if (rxstat & RE_RDESC_STAT_CRCERR)
   1218 				aprint_error(", CRC error");
   1219 			aprint_error("\n");
   1220 #endif
   1221 			ifp->if_ierrors++;
   1222 			/*
   1223 			 * If this is part of a multi-fragment packet,
   1224 			 * discard all the pieces.
   1225 			 */
   1226 			if (sc->re_head != NULL) {
   1227 				m_freem(sc->re_head);
   1228 				sc->re_head = sc->re_tail = NULL;
   1229 			}
   1230 			re_newbuf(sc, i, m);
   1231 			continue;
   1232 		}
   1233 
   1234 		/*
   1235 		 * If allocating a replacement mbuf fails,
   1236 		 * reload the current one.
   1237 		 */
   1238 
   1239 		if (__predict_false(re_newbuf(sc, i, NULL) != 0)) {
   1240 			ifp->if_ierrors++;
   1241 			if (sc->re_head != NULL) {
   1242 				m_freem(sc->re_head);
   1243 				sc->re_head = sc->re_tail = NULL;
   1244 			}
   1245 			re_newbuf(sc, i, m);
   1246 			continue;
   1247 		}
   1248 
   1249 		if (sc->re_head != NULL) {
   1250 			m->m_len = total_len % (MCLBYTES - RE_ETHER_ALIGN);
   1251 			/*
   1252 			 * Special case: if there's 4 bytes or less
   1253 			 * in this buffer, the mbuf can be discarded:
   1254 			 * the last 4 bytes is the CRC, which we don't
   1255 			 * care about anyway.
   1256 			 */
   1257 			if (m->m_len <= ETHER_CRC_LEN) {
   1258 				sc->re_tail->m_len -=
   1259 				    (ETHER_CRC_LEN - m->m_len);
   1260 				m_freem(m);
   1261 			} else {
   1262 				m->m_len -= ETHER_CRC_LEN;
   1263 				m->m_flags &= ~M_PKTHDR;
   1264 				sc->re_tail->m_next = m;
   1265 			}
   1266 			m = sc->re_head;
   1267 			sc->re_head = sc->re_tail = NULL;
   1268 			m->m_pkthdr.len = total_len - ETHER_CRC_LEN;
   1269 		} else
   1270 			m->m_pkthdr.len = m->m_len =
   1271 			    (total_len - ETHER_CRC_LEN);
   1272 
   1273 		ifp->if_ipackets++;
   1274 		m->m_pkthdr.rcvif = ifp;
   1275 
   1276 		/* Do RX checksumming */
   1277 
   1278 		/* Check IP header checksum */
   1279 		if (rxstat & RE_RDESC_STAT_PROTOID) {
   1280 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
   1281 			if (rxstat & RE_RDESC_STAT_IPSUMBAD)
   1282 				m->m_pkthdr.csum_flags |= M_CSUM_IPv4_BAD;
   1283 		}
   1284 
   1285 		/* Check TCP/UDP checksum */
   1286 		if (RE_TCPPKT(rxstat)) {
   1287 			m->m_pkthdr.csum_flags |= M_CSUM_TCPv4;
   1288 			if (rxstat & RE_RDESC_STAT_TCPSUMBAD)
   1289 				m->m_pkthdr.csum_flags |= M_CSUM_TCP_UDP_BAD;
   1290 		} else if (RE_UDPPKT(rxstat)) {
   1291 			m->m_pkthdr.csum_flags |= M_CSUM_UDPv4;
   1292 			if (rxstat & RE_RDESC_STAT_UDPSUMBAD)
   1293 				m->m_pkthdr.csum_flags |= M_CSUM_TCP_UDP_BAD;
   1294 		}
   1295 
   1296 		if (rxvlan & RE_RDESC_VLANCTL_TAG) {
   1297 			VLAN_INPUT_TAG(ifp, m,
   1298 			     bswap16(rxvlan & RE_RDESC_VLANCTL_DATA),
   1299 			     continue);
   1300 		}
   1301 #if NBPFILTER > 0
   1302 		if (ifp->if_bpf)
   1303 			bpf_mtap(ifp->if_bpf, m);
   1304 #endif
   1305 		(*ifp->if_input)(ifp, m);
   1306 	}
   1307 
   1308 	sc->re_ldata.re_rx_prodidx = i;
   1309 }
   1310 
   1311 static void
   1312 re_txeof(struct rtk_softc *sc)
   1313 {
   1314 	struct ifnet		*ifp;
   1315 	struct re_txq		*txq;
   1316 	uint32_t		txstat;
   1317 	int			idx, descidx;
   1318 
   1319 	ifp = &sc->ethercom.ec_if;
   1320 
   1321 	for (idx = sc->re_ldata.re_txq_considx;
   1322 	    sc->re_ldata.re_txq_free < RE_TX_QLEN;
   1323 	    idx = RE_NEXT_TXQ(sc, idx), sc->re_ldata.re_txq_free++) {
   1324 		txq = &sc->re_ldata.re_txq[idx];
   1325 		KASSERT(txq->txq_mbuf != NULL);
   1326 
   1327 		descidx = txq->txq_descidx;
   1328 		RE_TXDESCSYNC(sc, descidx,
   1329 		    BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
   1330 		txstat =
   1331 		    le32toh(sc->re_ldata.re_tx_list[descidx].re_cmdstat);
   1332 		RE_TXDESCSYNC(sc, descidx, BUS_DMASYNC_PREREAD);
   1333 		KASSERT((txstat & RE_TDESC_CMD_EOF) != 0);
   1334 		if (txstat & RE_TDESC_CMD_OWN) {
   1335 			break;
   1336 		}
   1337 
   1338 		sc->re_ldata.re_tx_free += txq->txq_nsegs;
   1339 		KASSERT(sc->re_ldata.re_tx_free <= RE_TX_DESC_CNT(sc));
   1340 		bus_dmamap_sync(sc->sc_dmat, txq->txq_dmamap,
   1341 		    0, txq->txq_dmamap->dm_mapsize, BUS_DMASYNC_POSTWRITE);
   1342 		bus_dmamap_unload(sc->sc_dmat, txq->txq_dmamap);
   1343 		m_freem(txq->txq_mbuf);
   1344 		txq->txq_mbuf = NULL;
   1345 
   1346 		if (txstat & (RE_TDESC_STAT_EXCESSCOL | RE_TDESC_STAT_COLCNT))
   1347 			ifp->if_collisions++;
   1348 		if (txstat & RE_TDESC_STAT_TXERRSUM)
   1349 			ifp->if_oerrors++;
   1350 		else
   1351 			ifp->if_opackets++;
   1352 	}
   1353 
   1354 	sc->re_ldata.re_txq_considx = idx;
   1355 
   1356 	if (sc->re_ldata.re_txq_free > RE_NTXDESC_RSVD)
   1357 		ifp->if_flags &= ~IFF_OACTIVE;
   1358 
   1359 	/*
   1360 	 * If not all descriptors have been released reaped yet,
   1361 	 * reload the timer so that we will eventually get another
   1362 	 * interrupt that will cause us to re-enter this routine.
   1363 	 * This is done in case the transmitter has gone idle.
   1364 	 */
   1365 	if (sc->re_ldata.re_txq_free < RE_TX_QLEN) {
   1366 		CSR_WRITE_4(sc, RTK_TIMERCNT, 1);
   1367 		if ((sc->sc_quirk & RTKQ_PCIE) != 0) {
   1368 			/*
   1369 			 * Some chips will ignore a second TX request
   1370 			 * issued while an existing transmission is in
   1371 			 * progress. If the transmitter goes idle but
   1372 			 * there are still packets waiting to be sent,
   1373 			 * we need to restart the channel here to flush
   1374 			 * them out. This only seems to be required with
   1375 			 * the PCIe devices.
   1376 			 */
   1377 			CSR_WRITE_1(sc, RTK_GTXSTART, RTK_TXSTART_START);
   1378 		}
   1379 	} else
   1380 		ifp->if_timer = 0;
   1381 }
   1382 
   1383 /*
   1384  * Stop all chip I/O so that the kernel's probe routines don't
   1385  * get confused by errant DMAs when rebooting.
   1386  */
   1387 static void
   1388 re_shutdown(void *vsc)
   1389 
   1390 {
   1391 	struct rtk_softc	*sc = vsc;
   1392 
   1393 	re_stop(&sc->ethercom.ec_if, 0);
   1394 }
   1395 
   1396 
   1397 static void
   1398 re_tick(void *xsc)
   1399 {
   1400 	struct rtk_softc	*sc = xsc;
   1401 	int s;
   1402 
   1403 	/*XXX: just return for 8169S/8110S with rev 2 or newer phy */
   1404 	s = splnet();
   1405 
   1406 	mii_tick(&sc->mii);
   1407 	splx(s);
   1408 
   1409 	callout_reset(&sc->rtk_tick_ch, hz, re_tick, sc);
   1410 }
   1411 
   1412 #ifdef DEVICE_POLLING
   1413 static void
   1414 re_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
   1415 {
   1416 	struct rtk_softc *sc = ifp->if_softc;
   1417 
   1418 	RTK_LOCK(sc);
   1419 	if ((ifp->if_capenable & IFCAP_POLLING) == 0) {
   1420 		ether_poll_deregister(ifp);
   1421 		cmd = POLL_DEREGISTER;
   1422 	}
   1423 	if (cmd == POLL_DEREGISTER) { /* final call, enable interrupts */
   1424 		CSR_WRITE_2(sc, RTK_IMR, RTK_INTRS_CPLUS);
   1425 		goto done;
   1426 	}
   1427 
   1428 	sc->rxcycles = count;
   1429 	re_rxeof(sc);
   1430 	re_txeof(sc);
   1431 
   1432 	if (IFQ_IS_EMPTY(&ifp->if_snd) == 0)
   1433 		(*ifp->if_start)(ifp);
   1434 
   1435 	if (cmd == POLL_AND_CHECK_STATUS) { /* also check status register */
   1436 		uint16_t       status;
   1437 
   1438 		status = CSR_READ_2(sc, RTK_ISR);
   1439 		if (status == 0xffff)
   1440 			goto done;
   1441 		if (status)
   1442 			CSR_WRITE_2(sc, RTK_ISR, status);
   1443 
   1444 		/*
   1445 		 * XXX check behaviour on receiver stalls.
   1446 		 */
   1447 
   1448 		if (status & RTK_ISR_SYSTEM_ERR) {
   1449 			re_init(sc);
   1450 		}
   1451 	}
   1452  done:
   1453 	RTK_UNLOCK(sc);
   1454 }
   1455 #endif /* DEVICE_POLLING */
   1456 
   1457 int
   1458 re_intr(void *arg)
   1459 {
   1460 	struct rtk_softc	*sc = arg;
   1461 	struct ifnet		*ifp;
   1462 	uint16_t		status;
   1463 	int			handled = 0;
   1464 
   1465 	ifp = &sc->ethercom.ec_if;
   1466 
   1467 	if ((ifp->if_flags & IFF_UP) == 0)
   1468 		return 0;
   1469 
   1470 #ifdef DEVICE_POLLING
   1471 	if (ifp->if_flags & IFF_POLLING)
   1472 		goto done;
   1473 	if ((ifp->if_capenable & IFCAP_POLLING) &&
   1474 	    ether_poll_register(re_poll, ifp)) { /* ok, disable interrupts */
   1475 		CSR_WRITE_2(sc, RTK_IMR, 0x0000);
   1476 		re_poll(ifp, 0, 1);
   1477 		goto done;
   1478 	}
   1479 #endif /* DEVICE_POLLING */
   1480 
   1481 	for (;;) {
   1482 
   1483 		status = CSR_READ_2(sc, RTK_ISR);
   1484 		/* If the card has gone away the read returns 0xffff. */
   1485 		if (status == 0xffff)
   1486 			break;
   1487 		if (status) {
   1488 			handled = 1;
   1489 			CSR_WRITE_2(sc, RTK_ISR, status);
   1490 		}
   1491 
   1492 		if ((status & RTK_INTRS_CPLUS) == 0)
   1493 			break;
   1494 
   1495 		if (status & (RTK_ISR_RX_OK | RTK_ISR_RX_ERR))
   1496 			re_rxeof(sc);
   1497 
   1498 		if (status & (RTK_ISR_TIMEOUT_EXPIRED | RTK_ISR_TX_ERR |
   1499 		    RTK_ISR_TX_DESC_UNAVAIL))
   1500 			re_txeof(sc);
   1501 
   1502 		if (status & RTK_ISR_SYSTEM_ERR) {
   1503 			re_init(ifp);
   1504 		}
   1505 
   1506 		if (status & RTK_ISR_LINKCHG) {
   1507 			callout_stop(&sc->rtk_tick_ch);
   1508 			re_tick(sc);
   1509 		}
   1510 	}
   1511 
   1512 	if (handled && !IFQ_IS_EMPTY(&ifp->if_snd))
   1513 		re_start(ifp);
   1514 
   1515 #ifdef DEVICE_POLLING
   1516  done:
   1517 #endif
   1518 
   1519 	return handled;
   1520 }
   1521 
   1522 
   1523 
   1524 /*
   1525  * Main transmit routine for C+ and gigE NICs.
   1526  */
   1527 
   1528 static void
   1529 re_start(struct ifnet *ifp)
   1530 {
   1531 	struct rtk_softc	*sc;
   1532 	struct mbuf		*m;
   1533 	bus_dmamap_t		map;
   1534 	struct re_txq		*txq;
   1535 	struct re_desc		*d;
   1536 	struct m_tag		*mtag;
   1537 	uint32_t		cmdstat, re_flags, vlanctl;
   1538 	int			ofree, idx, error, nsegs, seg;
   1539 	int			startdesc, curdesc, lastdesc;
   1540 	boolean_t		pad;
   1541 
   1542 	sc = ifp->if_softc;
   1543 	ofree = sc->re_ldata.re_txq_free;
   1544 
   1545 	for (idx = sc->re_ldata.re_txq_prodidx;; idx = RE_NEXT_TXQ(sc, idx)) {
   1546 
   1547 		IFQ_POLL(&ifp->if_snd, m);
   1548 		if (m == NULL)
   1549 			break;
   1550 
   1551 		if (sc->re_ldata.re_txq_free == 0 ||
   1552 		    sc->re_ldata.re_tx_free == 0) {
   1553 			/* no more free slots left */
   1554 			ifp->if_flags |= IFF_OACTIVE;
   1555 			break;
   1556 		}
   1557 
   1558 		/*
   1559 		 * Set up checksum offload. Note: checksum offload bits must
   1560 		 * appear in all descriptors of a multi-descriptor transmit
   1561 		 * attempt. (This is according to testing done with an 8169
   1562 		 * chip. I'm not sure if this is a requirement or a bug.)
   1563 		 */
   1564 
   1565 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
   1566 			uint32_t segsz = m->m_pkthdr.segsz;
   1567 
   1568 			re_flags = RE_TDESC_CMD_LGSEND |
   1569 			    (segsz << RE_TDESC_CMD_MSSVAL_SHIFT);
   1570 		} else {
   1571 			/*
   1572 			 * set RE_TDESC_CMD_IPCSUM if any checksum offloading
   1573 			 * is requested.  otherwise, RE_TDESC_CMD_TCPCSUM/
   1574 			 * RE_TDESC_CMD_UDPCSUM doesn't make effects.
   1575 			 */
   1576 			re_flags = 0;
   1577 			if ((m->m_pkthdr.csum_flags &
   1578 			    (M_CSUM_IPv4 | M_CSUM_TCPv4 | M_CSUM_UDPv4))
   1579 			    != 0) {
   1580 				re_flags |= RE_TDESC_CMD_IPCSUM;
   1581 				if (m->m_pkthdr.csum_flags & M_CSUM_TCPv4) {
   1582 					re_flags |= RE_TDESC_CMD_TCPCSUM;
   1583 				} else if (m->m_pkthdr.csum_flags &
   1584 				    M_CSUM_UDPv4) {
   1585 					re_flags |= RE_TDESC_CMD_UDPCSUM;
   1586 				}
   1587 			}
   1588 		}
   1589 
   1590 		txq = &sc->re_ldata.re_txq[idx];
   1591 		map = txq->txq_dmamap;
   1592 		error = bus_dmamap_load_mbuf(sc->sc_dmat, map, m,
   1593 		    BUS_DMA_WRITE|BUS_DMA_NOWAIT);
   1594 
   1595 		if (__predict_false(error)) {
   1596 			/* XXX try to defrag if EFBIG? */
   1597 			aprint_error("%s: can't map mbuf (error %d)\n",
   1598 			    sc->sc_dev.dv_xname, error);
   1599 
   1600 			IFQ_DEQUEUE(&ifp->if_snd, m);
   1601 			m_freem(m);
   1602 			ifp->if_oerrors++;
   1603 			continue;
   1604 		}
   1605 
   1606 		nsegs = map->dm_nsegs;
   1607 		pad = FALSE;
   1608 		if (__predict_false(m->m_pkthdr.len <= RE_IP4CSUMTX_PADLEN &&
   1609 		    (re_flags & RE_TDESC_CMD_IPCSUM) != 0)) {
   1610 			pad = TRUE;
   1611 			nsegs++;
   1612 		}
   1613 
   1614 		if (nsegs > sc->re_ldata.re_tx_free) {
   1615 			/*
   1616 			 * Not enough free descriptors to transmit this packet.
   1617 			 */
   1618 			ifp->if_flags |= IFF_OACTIVE;
   1619 			bus_dmamap_unload(sc->sc_dmat, map);
   1620 			break;
   1621 		}
   1622 
   1623 		IFQ_DEQUEUE(&ifp->if_snd, m);
   1624 
   1625 		/*
   1626 		 * Make sure that the caches are synchronized before we
   1627 		 * ask the chip to start DMA for the packet data.
   1628 		 */
   1629 		bus_dmamap_sync(sc->sc_dmat, map, 0, map->dm_mapsize,
   1630 		    BUS_DMASYNC_PREWRITE);
   1631 
   1632 		/*
   1633 		 * Set up hardware VLAN tagging. Note: vlan tag info must
   1634 		 * appear in all descriptors of a multi-descriptor
   1635 		 * transmission attempt.
   1636 		 */
   1637 		vlanctl = 0;
   1638 		if ((mtag = VLAN_OUTPUT_TAG(&sc->ethercom, m)) != NULL)
   1639 			vlanctl = bswap16(VLAN_TAG_VALUE(mtag)) |
   1640 			    RE_TDESC_VLANCTL_TAG;
   1641 
   1642 		/*
   1643 		 * Map the segment array into descriptors.
   1644 		 * Note that we set the start-of-frame and
   1645 		 * end-of-frame markers for either TX or RX,
   1646 		 * but they really only have meaning in the TX case.
   1647 		 * (In the RX case, it's the chip that tells us
   1648 		 *  where packets begin and end.)
   1649 		 * We also keep track of the end of the ring
   1650 		 * and set the end-of-ring bits as needed,
   1651 		 * and we set the ownership bits in all except
   1652 		 * the very first descriptor. (The caller will
   1653 		 * set this descriptor later when it start
   1654 		 * transmission or reception.)
   1655 		 */
   1656 		curdesc = startdesc = sc->re_ldata.re_tx_nextfree;
   1657 		lastdesc = -1;
   1658 		for (seg = 0; seg < map->dm_nsegs;
   1659 		    seg++, curdesc = RE_NEXT_TX_DESC(sc, curdesc)) {
   1660 			d = &sc->re_ldata.re_tx_list[curdesc];
   1661 #ifdef DIAGNOSTIC
   1662 			RE_TXDESCSYNC(sc, curdesc,
   1663 			    BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
   1664 			cmdstat = le32toh(d->re_cmdstat);
   1665 			RE_TXDESCSYNC(sc, curdesc, BUS_DMASYNC_PREREAD);
   1666 			if (cmdstat & RE_TDESC_STAT_OWN) {
   1667 				panic("%s: tried to map busy TX descriptor",
   1668 				    sc->sc_dev.dv_xname);
   1669 			}
   1670 #endif
   1671 
   1672 			d->re_vlanctl = htole32(vlanctl);
   1673 			re_set_bufaddr(d, map->dm_segs[seg].ds_addr);
   1674 			cmdstat = re_flags | map->dm_segs[seg].ds_len;
   1675 			if (seg == 0)
   1676 				cmdstat |= RE_TDESC_CMD_SOF;
   1677 			else
   1678 				cmdstat |= RE_TDESC_CMD_OWN;
   1679 			if (curdesc == (RE_TX_DESC_CNT(sc) - 1))
   1680 				cmdstat |= RE_TDESC_CMD_EOR;
   1681 			if (seg == nsegs - 1) {
   1682 				cmdstat |= RE_TDESC_CMD_EOF;
   1683 				lastdesc = curdesc;
   1684 			}
   1685 			d->re_cmdstat = htole32(cmdstat);
   1686 			RE_TXDESCSYNC(sc, curdesc,
   1687 			    BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
   1688 		}
   1689 		if (__predict_false(pad)) {
   1690 			bus_addr_t paddaddr;
   1691 
   1692 			d = &sc->re_ldata.re_tx_list[curdesc];
   1693 			d->re_vlanctl = htole32(vlanctl);
   1694 			paddaddr = RE_TXPADDADDR(sc);
   1695 			re_set_bufaddr(d, paddaddr);
   1696 			cmdstat = re_flags |
   1697 			    RE_TDESC_CMD_OWN | RE_TDESC_CMD_EOF |
   1698 			    (RE_IP4CSUMTX_PADLEN + 1 - m->m_pkthdr.len);
   1699 			if (curdesc == (RE_TX_DESC_CNT(sc) - 1))
   1700 				cmdstat |= RE_TDESC_CMD_EOR;
   1701 			d->re_cmdstat = htole32(cmdstat);
   1702 			RE_TXDESCSYNC(sc, curdesc,
   1703 			    BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
   1704 			lastdesc = curdesc;
   1705 			curdesc = RE_NEXT_TX_DESC(sc, curdesc);
   1706 		}
   1707 		KASSERT(lastdesc != -1);
   1708 
   1709 		/* Transfer ownership of packet to the chip. */
   1710 
   1711 		sc->re_ldata.re_tx_list[startdesc].re_cmdstat |=
   1712 		    htole32(RE_TDESC_CMD_OWN);
   1713 		RE_TXDESCSYNC(sc, startdesc,
   1714 		    BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
   1715 
   1716 		/* update info of TX queue and descriptors */
   1717 		txq->txq_mbuf = m;
   1718 		txq->txq_descidx = lastdesc;
   1719 		txq->txq_nsegs = nsegs;
   1720 
   1721 		sc->re_ldata.re_txq_free--;
   1722 		sc->re_ldata.re_tx_free -= nsegs;
   1723 		sc->re_ldata.re_tx_nextfree = curdesc;
   1724 
   1725 #if NBPFILTER > 0
   1726 		/*
   1727 		 * If there's a BPF listener, bounce a copy of this frame
   1728 		 * to him.
   1729 		 */
   1730 		if (ifp->if_bpf)
   1731 			bpf_mtap(ifp->if_bpf, m);
   1732 #endif
   1733 	}
   1734 
   1735 	if (sc->re_ldata.re_txq_free < ofree) {
   1736 		/*
   1737 		 * TX packets are enqueued.
   1738 		 */
   1739 		sc->re_ldata.re_txq_prodidx = idx;
   1740 
   1741 		/*
   1742 		 * Start the transmitter to poll.
   1743 		 *
   1744 		 * RealTek put the TX poll request register in a different
   1745 		 * location on the 8169 gigE chip. I don't know why.
   1746 		 */
   1747 		if ((sc->sc_quirk & RTKQ_8139CPLUS) != 0)
   1748 			CSR_WRITE_1(sc, RTK_TXSTART, RTK_TXSTART_START);
   1749 		else
   1750 			CSR_WRITE_1(sc, RTK_GTXSTART, RTK_TXSTART_START);
   1751 
   1752 		/*
   1753 		 * Use the countdown timer for interrupt moderation.
   1754 		 * 'TX done' interrupts are disabled. Instead, we reset the
   1755 		 * countdown timer, which will begin counting until it hits
   1756 		 * the value in the TIMERINT register, and then trigger an
   1757 		 * interrupt. Each time we write to the TIMERCNT register,
   1758 		 * the timer count is reset to 0.
   1759 		 */
   1760 		CSR_WRITE_4(sc, RTK_TIMERCNT, 1);
   1761 
   1762 		/*
   1763 		 * Set a timeout in case the chip goes out to lunch.
   1764 		 */
   1765 		ifp->if_timer = 5;
   1766 	}
   1767 }
   1768 
   1769 static int
   1770 re_init(struct ifnet *ifp)
   1771 {
   1772 	struct rtk_softc	*sc = ifp->if_softc;
   1773 	uint8_t			*enaddr;
   1774 	uint32_t		rxcfg = 0;
   1775 	uint32_t		reg;
   1776 	int error;
   1777 
   1778 	if ((error = re_enable(sc)) != 0)
   1779 		goto out;
   1780 
   1781 	/*
   1782 	 * Cancel pending I/O and free all RX/TX buffers.
   1783 	 */
   1784 	re_stop(ifp, 0);
   1785 
   1786 	re_reset(sc);
   1787 
   1788 	/*
   1789 	 * Enable C+ RX and TX mode, as well as VLAN stripping and
   1790 	 * RX checksum offload. We must configure the C+ register
   1791 	 * before all others.
   1792 	 */
   1793 	reg = 0;
   1794 
   1795 	/*
   1796 	 * XXX: Realtek docs say bits 0 and 1 are reserved, for 8169S/8110S.
   1797 	 * FreeBSD  drivers set these bits anyway (for 8139C+?).
   1798 	 * So far, it works.
   1799 	 */
   1800 
   1801 	/*
   1802 	 * XXX: For old 8169 set bit 14.
   1803 	 *      For 8169S/8110S and above, do not set bit 14.
   1804 	 */
   1805 	if ((sc->sc_quirk & RTKQ_8169NONS) != 0)
   1806 		reg |= (0x1 << 14) | RTK_CPLUSCMD_PCI_MRW;;
   1807 
   1808 	if (1)  {/* not for 8169S ? */
   1809 		reg |=
   1810 		    RTK_CPLUSCMD_VLANSTRIP |
   1811 		    (ifp->if_capenable &
   1812 		    (IFCAP_CSUM_IPv4_Rx | IFCAP_CSUM_TCPv4_Rx |
   1813 		     IFCAP_CSUM_UDPv4_Rx) ?
   1814 		    RTK_CPLUSCMD_RXCSUM_ENB : 0);
   1815 	}
   1816 
   1817 	CSR_WRITE_2(sc, RTK_CPLUS_CMD,
   1818 	    reg | RTK_CPLUSCMD_RXENB | RTK_CPLUSCMD_TXENB);
   1819 
   1820 	/* XXX: from Realtek-supplied Linux driver. Wholly undocumented. */
   1821 	if ((sc->sc_quirk & RTKQ_8139CPLUS) == 0)
   1822 		CSR_WRITE_2(sc, RTK_IM, 0x0000);
   1823 
   1824 	DELAY(10000);
   1825 
   1826 	/*
   1827 	 * Init our MAC address.  Even though the chipset
   1828 	 * documentation doesn't mention it, we need to enter "Config
   1829 	 * register write enable" mode to modify the ID registers.
   1830 	 */
   1831 	CSR_WRITE_1(sc, RTK_EECMD, RTK_EEMODE_WRITECFG);
   1832 	enaddr = LLADDR(ifp->if_sadl);
   1833 	reg = enaddr[0] | (enaddr[1] << 8) |
   1834 	    (enaddr[2] << 16) | (enaddr[3] << 24);
   1835 	CSR_WRITE_4(sc, RTK_IDR0, reg);
   1836 	reg = enaddr[4] | (enaddr[5] << 8);
   1837 	CSR_WRITE_4(sc, RTK_IDR4, reg);
   1838 	CSR_WRITE_1(sc, RTK_EECMD, RTK_EEMODE_OFF);
   1839 
   1840 	/*
   1841 	 * For C+ mode, initialize the RX descriptors and mbufs.
   1842 	 */
   1843 	re_rx_list_init(sc);
   1844 	re_tx_list_init(sc);
   1845 
   1846 	/*
   1847 	 * Load the addresses of the RX and TX lists into the chip.
   1848 	 */
   1849 	CSR_WRITE_4(sc, RTK_RXLIST_ADDR_HI,
   1850 	    RE_ADDR_HI(sc->re_ldata.re_rx_list_map->dm_segs[0].ds_addr));
   1851 	CSR_WRITE_4(sc, RTK_RXLIST_ADDR_LO,
   1852 	    RE_ADDR_LO(sc->re_ldata.re_rx_list_map->dm_segs[0].ds_addr));
   1853 
   1854 	CSR_WRITE_4(sc, RTK_TXLIST_ADDR_HI,
   1855 	    RE_ADDR_HI(sc->re_ldata.re_tx_list_map->dm_segs[0].ds_addr));
   1856 	CSR_WRITE_4(sc, RTK_TXLIST_ADDR_LO,
   1857 	    RE_ADDR_LO(sc->re_ldata.re_tx_list_map->dm_segs[0].ds_addr));
   1858 
   1859 	/*
   1860 	 * Enable transmit and receive.
   1861 	 */
   1862 	CSR_WRITE_1(sc, RTK_COMMAND, RTK_CMD_TX_ENB | RTK_CMD_RX_ENB);
   1863 
   1864 	/*
   1865 	 * Set the initial TX and RX configuration.
   1866 	 */
   1867 	if (sc->re_testmode && (sc->sc_quirk & RTKQ_8169NONS) != 0) {
   1868 		/* test mode is needed only for old 8169 */
   1869 		CSR_WRITE_4(sc, RTK_TXCFG,
   1870 		    RE_TXCFG_CONFIG | RTK_LOOPTEST_ON);
   1871 	} else
   1872 		CSR_WRITE_4(sc, RTK_TXCFG, RE_TXCFG_CONFIG);
   1873 
   1874 	CSR_WRITE_1(sc, RTK_EARLY_TX_THRESH, 16);
   1875 
   1876 	CSR_WRITE_4(sc, RTK_RXCFG, RE_RXCFG_CONFIG);
   1877 
   1878 	/* Set the individual bit to receive frames for this host only. */
   1879 	rxcfg = CSR_READ_4(sc, RTK_RXCFG);
   1880 	rxcfg |= RTK_RXCFG_RX_INDIV;
   1881 
   1882 	/* If we want promiscuous mode, set the allframes bit. */
   1883 	if (ifp->if_flags & IFF_PROMISC)
   1884 		rxcfg |= RTK_RXCFG_RX_ALLPHYS;
   1885 	else
   1886 		rxcfg &= ~RTK_RXCFG_RX_ALLPHYS;
   1887 	CSR_WRITE_4(sc, RTK_RXCFG, rxcfg);
   1888 
   1889 	/*
   1890 	 * Set capture broadcast bit to capture broadcast frames.
   1891 	 */
   1892 	if (ifp->if_flags & IFF_BROADCAST)
   1893 		rxcfg |= RTK_RXCFG_RX_BROAD;
   1894 	else
   1895 		rxcfg &= ~RTK_RXCFG_RX_BROAD;
   1896 	CSR_WRITE_4(sc, RTK_RXCFG, rxcfg);
   1897 
   1898 	/*
   1899 	 * Program the multicast filter, if necessary.
   1900 	 */
   1901 	rtk_setmulti(sc);
   1902 
   1903 #ifdef DEVICE_POLLING
   1904 	/*
   1905 	 * Disable interrupts if we are polling.
   1906 	 */
   1907 	if (ifp->if_flags & IFF_POLLING)
   1908 		CSR_WRITE_2(sc, RTK_IMR, 0);
   1909 	else	/* otherwise ... */
   1910 #endif /* DEVICE_POLLING */
   1911 	/*
   1912 	 * Enable interrupts.
   1913 	 */
   1914 	if (sc->re_testmode)
   1915 		CSR_WRITE_2(sc, RTK_IMR, 0);
   1916 	else
   1917 		CSR_WRITE_2(sc, RTK_IMR, RTK_INTRS_CPLUS);
   1918 
   1919 	/* Start RX/TX process. */
   1920 	CSR_WRITE_4(sc, RTK_MISSEDPKT, 0);
   1921 #ifdef notdef
   1922 	/* Enable receiver and transmitter. */
   1923 	CSR_WRITE_1(sc, RTK_COMMAND, RTK_CMD_TX_ENB | RTK_CMD_RX_ENB);
   1924 #endif
   1925 
   1926 	/*
   1927 	 * Initialize the timer interrupt register so that
   1928 	 * a timer interrupt will be generated once the timer
   1929 	 * reaches a certain number of ticks. The timer is
   1930 	 * reloaded on each transmit. This gives us TX interrupt
   1931 	 * moderation, which dramatically improves TX frame rate.
   1932 	 */
   1933 
   1934 	if ((sc->sc_quirk & RTKQ_8139CPLUS) != 0)
   1935 		CSR_WRITE_4(sc, RTK_TIMERINT, 0x400);
   1936 	else {
   1937 		CSR_WRITE_4(sc, RTK_TIMERINT_8169, 0x800);
   1938 
   1939 		/*
   1940 		 * For 8169 gigE NICs, set the max allowed RX packet
   1941 		 * size so we can receive jumbo frames.
   1942 		 */
   1943 		CSR_WRITE_2(sc, RTK_MAXRXPKTLEN, 16383);
   1944 	}
   1945 
   1946 	if (sc->re_testmode)
   1947 		return 0;
   1948 
   1949 	CSR_WRITE_1(sc, RTK_CFG1, RTK_CFG1_DRVLOAD);
   1950 
   1951 	ifp->if_flags |= IFF_RUNNING;
   1952 	ifp->if_flags &= ~IFF_OACTIVE;
   1953 
   1954 	callout_reset(&sc->rtk_tick_ch, hz, re_tick, sc);
   1955 
   1956  out:
   1957 	if (error) {
   1958 		ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
   1959 		ifp->if_timer = 0;
   1960 		aprint_error("%s: interface not running\n",
   1961 		    sc->sc_dev.dv_xname);
   1962 	}
   1963 
   1964 	return error;
   1965 }
   1966 
   1967 /*
   1968  * Set media options.
   1969  */
   1970 static int
   1971 re_ifmedia_upd(struct ifnet *ifp)
   1972 {
   1973 	struct rtk_softc	*sc;
   1974 
   1975 	sc = ifp->if_softc;
   1976 
   1977 	return mii_mediachg(&sc->mii);
   1978 }
   1979 
   1980 /*
   1981  * Report current media status.
   1982  */
   1983 static void
   1984 re_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
   1985 {
   1986 	struct rtk_softc	*sc;
   1987 
   1988 	sc = ifp->if_softc;
   1989 
   1990 	mii_pollstat(&sc->mii);
   1991 	ifmr->ifm_active = sc->mii.mii_media_active;
   1992 	ifmr->ifm_status = sc->mii.mii_media_status;
   1993 }
   1994 
   1995 static int
   1996 re_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
   1997 {
   1998 	struct rtk_softc	*sc = ifp->if_softc;
   1999 	struct ifreq		*ifr = (struct ifreq *) data;
   2000 	int			s, error = 0;
   2001 
   2002 	s = splnet();
   2003 
   2004 	switch (command) {
   2005 	case SIOCSIFMTU:
   2006 		if (ifr->ifr_mtu > RE_JUMBO_MTU)
   2007 			error = EINVAL;
   2008 		ifp->if_mtu = ifr->ifr_mtu;
   2009 		break;
   2010 	case SIOCGIFMEDIA:
   2011 	case SIOCSIFMEDIA:
   2012 		error = ifmedia_ioctl(ifp, ifr, &sc->mii.mii_media, command);
   2013 		break;
   2014 	default:
   2015 		error = ether_ioctl(ifp, command, data);
   2016 		if (error == ENETRESET) {
   2017 			if (ifp->if_flags & IFF_RUNNING)
   2018 				rtk_setmulti(sc);
   2019 			error = 0;
   2020 		}
   2021 		break;
   2022 	}
   2023 
   2024 	splx(s);
   2025 
   2026 	return error;
   2027 }
   2028 
   2029 static void
   2030 re_watchdog(struct ifnet *ifp)
   2031 {
   2032 	struct rtk_softc	*sc;
   2033 	int			s;
   2034 
   2035 	sc = ifp->if_softc;
   2036 	s = splnet();
   2037 	aprint_error("%s: watchdog timeout\n", sc->sc_dev.dv_xname);
   2038 	ifp->if_oerrors++;
   2039 
   2040 	re_txeof(sc);
   2041 	re_rxeof(sc);
   2042 
   2043 	re_init(ifp);
   2044 
   2045 	splx(s);
   2046 }
   2047 
   2048 /*
   2049  * Stop the adapter and free any mbufs allocated to the
   2050  * RX and TX lists.
   2051  */
   2052 static void
   2053 re_stop(struct ifnet *ifp, int disable)
   2054 {
   2055 	int		i;
   2056 	struct rtk_softc *sc = ifp->if_softc;
   2057 
   2058 	callout_stop(&sc->rtk_tick_ch);
   2059 
   2060 #ifdef DEVICE_POLLING
   2061 	ether_poll_deregister(ifp);
   2062 #endif /* DEVICE_POLLING */
   2063 
   2064 	mii_down(&sc->mii);
   2065 
   2066 	CSR_WRITE_1(sc, RTK_COMMAND, 0x00);
   2067 	CSR_WRITE_2(sc, RTK_IMR, 0x0000);
   2068 
   2069 	if (sc->re_head != NULL) {
   2070 		m_freem(sc->re_head);
   2071 		sc->re_head = sc->re_tail = NULL;
   2072 	}
   2073 
   2074 	/* Free the TX list buffers. */
   2075 	for (i = 0; i < RE_TX_QLEN; i++) {
   2076 		if (sc->re_ldata.re_txq[i].txq_mbuf != NULL) {
   2077 			bus_dmamap_unload(sc->sc_dmat,
   2078 			    sc->re_ldata.re_txq[i].txq_dmamap);
   2079 			m_freem(sc->re_ldata.re_txq[i].txq_mbuf);
   2080 			sc->re_ldata.re_txq[i].txq_mbuf = NULL;
   2081 		}
   2082 	}
   2083 
   2084 	/* Free the RX list buffers. */
   2085 	for (i = 0; i < RE_RX_DESC_CNT; i++) {
   2086 		if (sc->re_ldata.re_rxsoft[i].rxs_mbuf != NULL) {
   2087 			bus_dmamap_unload(sc->sc_dmat,
   2088 			    sc->re_ldata.re_rxsoft[i].rxs_dmamap);
   2089 			m_freem(sc->re_ldata.re_rxsoft[i].rxs_mbuf);
   2090 			sc->re_ldata.re_rxsoft[i].rxs_mbuf = NULL;
   2091 		}
   2092 	}
   2093 
   2094 	if (disable)
   2095 		re_disable(sc);
   2096 
   2097 	ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
   2098 	ifp->if_timer = 0;
   2099 }
   2100