freebsd-skq/sys/dev/le/lance.c

824 lines
19 KiB
C
Raw Normal View History

/* $NetBSD: lance.c,v 1.34 2005/12/24 20:27:30 perry Exp $ */
/*-
* Copyright (c) 1997, 1998 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Charles M. Hannum and by Jason R. Thorpe of the Numerical Aerospace
* Simulation Facility, NASA Ames Research Center.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/*-
* Copyright (c) 1992, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Ralph Campbell and Rick Macklem.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)if_le.c 8.2 (Berkeley) 11/16/93
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/bus.h>
#include <sys/endian.h>
#include <sys/lock.h>
#include <sys/kernel.h>
#include <sys/mbuf.h>
#include <sys/mutex.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_vlan_var.h>
- Revert if_le_pci.c rev. 1.2; although lnc(4) is now gone, le_pci_probe() still should return BUS_PROBE_LOW_PRIORITY instead of BUS_PROBE_DEFAULT in order to give pcn(4) a chance to attach in case it probes after le(4). - Rearrange the code related to RX interrupt handling so that ownership of RX descriptors is immediately returned to the NIC after we have copied the data of the hardware, allowing the NIC to already reuse the descriptor while we are processing the data in ifp->if_input(). This results in a small but measurable increase in RX throughput. As a side-effect, this moves the workaround for the LANCE revision C bug to am7900.c (still off by default as I doubt we will actually encounter such an old chip in a machine running FreeBSD) and the workaround for the bug in the VMware PCnet-PCI emulation to am79000.c, which is now also only compiled on i386 (resulting in a small increase in RX throughput on the other platforms). - Change the RX interrupt handlers so that the descriptor error bits are only check once in case there was no error instead of twice (inspired by the NetBSD pcn(4), which additionally predicts the error branch as false). - Fix the debugging output of the RX and TX interrupt handlers; while looping through the descriptors print info about the currently processed one instead of always the previously last used one; remove pointless printing of info about the RX descriptor bits after their values were reset. - Create the DMA tags used to allocate the memory for the init block, descriptors and packet buffers with the alignment the respective NIC actually requires rather than using PAGE_SIZE unconditionally. This might as well fix the alignment of the memory as it seems we do not inherit the alignment constraint from the parent DMA tag. - For the PCI variants double the number of RX descriptors and buffers from 8 to 16 as this minimizes the number of RX overflows im seeing with one NIC-mainboard combination. Nevertheless move reporting of overflows under debugging as they seem unavoidable with some crappy hardware. - Set the software style of the PCI variants to ILACC rather than PCnet-PCI as the former is was am79000.c actually implements. Should not make a difference for this driver though. - Fix the driver name part in the MODULE_DEPEND of the PCI front-end for ether. - Use different device descriptions for PCnet-Home and PCnet-PCI. - Fix some 0/NULL confusion in lance_get(). - Use bus_addr_t for sc_addr and bus_size_t for sc_memsize as these are more appropriate than u_long for these. - Remove the unused LE_DRIVER_NAME macro. - Add a comment describing why we are taking the LE_HTOLE* etc approach instead of using byteorder(9) functions directly. - Improve some comments and fix some wording. MFC after: 2 weeks
2006-05-16 21:04:01 +00:00
#include <machine/bus.h>
#include <dev/le/lancereg.h>
#include <dev/le/lancevar.h>
devclass_t le_devclass;
static void lance_start(struct ifnet *);
static void lance_stop(struct lance_softc *);
static void lance_init(void *);
static void lance_watchdog(void *s);
static int lance_mediachange(struct ifnet *);
static void lance_mediastatus(struct ifnet *, struct ifmediareq *);
static int lance_ioctl(struct ifnet *, u_long, caddr_t);
int
lance_config(struct lance_softc *sc, const char* name, int unit)
{
struct ifnet *ifp;
int i, nbuf;
if (LE_LOCK_INITIALIZED(sc) == 0)
return (ENXIO);
ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL)
return (ENOSPC);
callout_init_mtx(&sc->sc_wdog_ch, &sc->sc_mtx, 0);
/* Initialize ifnet structure. */
ifp->if_softc = sc;
if_initname(ifp, name, unit);
ifp->if_start = lance_start;
ifp->if_ioctl = lance_ioctl;
ifp->if_init = lance_init;
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
#ifdef LANCE_REVC_BUG
ifp->if_flags &= ~IFF_MULTICAST;
#endif
ifp->if_baudrate = IF_Mbps(10);
IFQ_SET_MAXLEN(&ifp->if_snd, IFQ_MAXLEN);
ifp->if_snd.ifq_drv_maxlen = IFQ_MAXLEN;
IFQ_SET_READY(&ifp->if_snd);
/* Initialize ifmedia structures. */
ifmedia_init(&sc->sc_media, 0, lance_mediachange, lance_mediastatus);
if (sc->sc_supmedia != NULL) {
for (i = 0; i < sc->sc_nsupmedia; i++)
ifmedia_add(&sc->sc_media, sc->sc_supmedia[i], 0, NULL);
ifmedia_set(&sc->sc_media, sc->sc_defaultmedia);
} else {
ifmedia_add(&sc->sc_media,
IFM_MAKEWORD(IFM_ETHER, IFM_MANUAL, 0, 0), 0, NULL);
ifmedia_set(&sc->sc_media,
IFM_MAKEWORD(IFM_ETHER, IFM_MANUAL, 0, 0));
}
switch (sc->sc_memsize) {
case 8192:
sc->sc_nrbuf = 4;
sc->sc_ntbuf = 1;
break;
case 16384:
sc->sc_nrbuf = 8;
sc->sc_ntbuf = 2;
break;
case 32768:
sc->sc_nrbuf = 16;
sc->sc_ntbuf = 4;
break;
case 65536:
sc->sc_nrbuf = 32;
sc->sc_ntbuf = 8;
break;
case 131072:
sc->sc_nrbuf = 64;
sc->sc_ntbuf = 16;
break;
case 262144:
sc->sc_nrbuf = 128;
sc->sc_ntbuf = 32;
break;
default:
/* weird memory size; cope with it */
nbuf = sc->sc_memsize / LEBLEN;
sc->sc_ntbuf = nbuf / 5;
sc->sc_nrbuf = nbuf - sc->sc_ntbuf;
}
if_printf(ifp, "%d receive buffers, %d transmit buffers\n",
sc->sc_nrbuf, sc->sc_ntbuf);
/* Make sure the chip is stopped. */
LE_LOCK(sc);
lance_stop(sc);
LE_UNLOCK(sc);
return (0);
}
void
lance_attach(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
/* Attach the interface. */
ether_ifattach(ifp, sc->sc_enaddr);
/* Claim 802.1q capability. */
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
ifp->if_capabilities |= IFCAP_VLAN_MTU;
ifp->if_capenable |= IFCAP_VLAN_MTU;
}
void
lance_detach(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
LE_LOCK(sc);
lance_stop(sc);
LE_UNLOCK(sc);
callout_drain(&sc->sc_wdog_ch);
ether_ifdetach(ifp);
if_free(ifp);
}
void
lance_suspend(struct lance_softc *sc)
{
LE_LOCK(sc);
lance_stop(sc);
LE_UNLOCK(sc);
}
void
lance_resume(struct lance_softc *sc)
{
LE_LOCK(sc);
if (sc->sc_ifp->if_flags & IFF_UP)
lance_init_locked(sc);
LE_UNLOCK(sc);
}
static void
lance_start(struct ifnet *ifp)
{
struct lance_softc *sc = ifp->if_softc;
LE_LOCK(sc);
(*sc->sc_start_locked)(sc);
LE_UNLOCK(sc);
}
static void
lance_stop(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
LE_LOCK_ASSERT(sc, MA_OWNED);
/*
* Mark the interface down and cancel the watchdog timer.
*/
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
callout_stop(&sc->sc_wdog_ch);
sc->sc_wdog_timer = 0;
(*sc->sc_wrcsr)(sc, LE_CSR0, LE_C0_STOP);
}
static void
lance_init(void *xsc)
{
struct lance_softc *sc = (struct lance_softc *)xsc;
LE_LOCK(sc);
lance_init_locked(sc);
LE_UNLOCK(sc);
}
/*
* Initialization of interface; set up initialization block
* and transmit/receive descriptor rings.
*/
void
lance_init_locked(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
u_long a;
int timo;
LE_LOCK_ASSERT(sc, MA_OWNED);
(*sc->sc_wrcsr)(sc, LE_CSR0, LE_C0_STOP);
DELAY(100);
/* Newer LANCE chips have a reset register. */
if (sc->sc_hwreset)
(*sc->sc_hwreset)(sc);
/* Set the correct byte swapping mode, etc. */
(*sc->sc_wrcsr)(sc, LE_CSR3, sc->sc_conf3);
/* Set the current media. This may require the chip to be stopped. */
if (sc->sc_mediachange)
(void)(*sc->sc_mediachange)(sc);
/*
* Update our private copy of the Ethernet address.
* We NEED the copy so we can ensure its alignment!
*/
memcpy(sc->sc_enaddr, IF_LLADDR(ifp), ETHER_ADDR_LEN);
/* Set up LANCE init block. */
(*sc->sc_meminit)(sc);
/* Give LANCE the physical address of its init block. */
a = sc->sc_addr + LE_INITADDR(sc);
(*sc->sc_wrcsr)(sc, LE_CSR1, a & 0xffff);
(*sc->sc_wrcsr)(sc, LE_CSR2, a >> 16);
/* Try to initialize the LANCE. */
DELAY(100);
(*sc->sc_wrcsr)(sc, LE_CSR0, LE_C0_INIT);
/* Wait for initialization to finish. */
for (timo = 100000; timo; timo--)
if ((*sc->sc_rdcsr)(sc, LE_CSR0) & LE_C0_IDON)
break;
if ((*sc->sc_rdcsr)(sc, LE_CSR0) & LE_C0_IDON) {
/* Start the LANCE. */
(*sc->sc_wrcsr)(sc, LE_CSR0, LE_C0_INEA | LE_C0_STRT);
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
sc->sc_wdog_timer = 0;
callout_reset(&sc->sc_wdog_ch, hz, lance_watchdog, sc);
(*sc->sc_start_locked)(sc);
} else
if_printf(ifp, "controller failed to initialize\n");
if (sc->sc_hwinit)
(*sc->sc_hwinit)(sc);
}
/*
* Routine to copy from mbuf chain to transmit buffer in
* network buffer memory.
*/
int
lance_put(struct lance_softc *sc, int boff, struct mbuf *m)
{
struct mbuf *n;
int len, tlen = 0;
LE_LOCK_ASSERT(sc, MA_OWNED);
for (; m; m = n) {
len = m->m_len;
if (len == 0) {
n = m_free(m);
m = NULL;
continue;
}
(*sc->sc_copytobuf)(sc, mtod(m, caddr_t), boff, len);
boff += len;
tlen += len;
n = m_free(m);
m = NULL;
}
if (tlen < LEMINSIZE) {
(*sc->sc_zerobuf)(sc, boff, LEMINSIZE - tlen);
tlen = LEMINSIZE;
}
return (tlen);
}
/*
* Pull data off an interface.
* Len is length of data, with local net header stripped.
* We copy the data into mbufs. When full cluster sized units are present
* we copy into clusters.
*/
- Revert if_le_pci.c rev. 1.2; although lnc(4) is now gone, le_pci_probe() still should return BUS_PROBE_LOW_PRIORITY instead of BUS_PROBE_DEFAULT in order to give pcn(4) a chance to attach in case it probes after le(4). - Rearrange the code related to RX interrupt handling so that ownership of RX descriptors is immediately returned to the NIC after we have copied the data of the hardware, allowing the NIC to already reuse the descriptor while we are processing the data in ifp->if_input(). This results in a small but measurable increase in RX throughput. As a side-effect, this moves the workaround for the LANCE revision C bug to am7900.c (still off by default as I doubt we will actually encounter such an old chip in a machine running FreeBSD) and the workaround for the bug in the VMware PCnet-PCI emulation to am79000.c, which is now also only compiled on i386 (resulting in a small increase in RX throughput on the other platforms). - Change the RX interrupt handlers so that the descriptor error bits are only check once in case there was no error instead of twice (inspired by the NetBSD pcn(4), which additionally predicts the error branch as false). - Fix the debugging output of the RX and TX interrupt handlers; while looping through the descriptors print info about the currently processed one instead of always the previously last used one; remove pointless printing of info about the RX descriptor bits after their values were reset. - Create the DMA tags used to allocate the memory for the init block, descriptors and packet buffers with the alignment the respective NIC actually requires rather than using PAGE_SIZE unconditionally. This might as well fix the alignment of the memory as it seems we do not inherit the alignment constraint from the parent DMA tag. - For the PCI variants double the number of RX descriptors and buffers from 8 to 16 as this minimizes the number of RX overflows im seeing with one NIC-mainboard combination. Nevertheless move reporting of overflows under debugging as they seem unavoidable with some crappy hardware. - Set the software style of the PCI variants to ILACC rather than PCnet-PCI as the former is was am79000.c actually implements. Should not make a difference for this driver though. - Fix the driver name part in the MODULE_DEPEND of the PCI front-end for ether. - Use different device descriptions for PCnet-Home and PCnet-PCI. - Fix some 0/NULL confusion in lance_get(). - Use bus_addr_t for sc_addr and bus_size_t for sc_memsize as these are more appropriate than u_long for these. - Remove the unused LE_DRIVER_NAME macro. - Add a comment describing why we are taking the LE_HTOLE* etc approach instead of using byteorder(9) functions directly. - Improve some comments and fix some wording. MFC after: 2 weeks
2006-05-16 21:04:01 +00:00
struct mbuf *
lance_get(struct lance_softc *sc, int boff, int totlen)
{
struct ifnet *ifp = sc->sc_ifp;
struct mbuf *m, *m0, *newm;
caddr_t newdata;
int len;
- Revert if_le_pci.c rev. 1.2; although lnc(4) is now gone, le_pci_probe() still should return BUS_PROBE_LOW_PRIORITY instead of BUS_PROBE_DEFAULT in order to give pcn(4) a chance to attach in case it probes after le(4). - Rearrange the code related to RX interrupt handling so that ownership of RX descriptors is immediately returned to the NIC after we have copied the data of the hardware, allowing the NIC to already reuse the descriptor while we are processing the data in ifp->if_input(). This results in a small but measurable increase in RX throughput. As a side-effect, this moves the workaround for the LANCE revision C bug to am7900.c (still off by default as I doubt we will actually encounter such an old chip in a machine running FreeBSD) and the workaround for the bug in the VMware PCnet-PCI emulation to am79000.c, which is now also only compiled on i386 (resulting in a small increase in RX throughput on the other platforms). - Change the RX interrupt handlers so that the descriptor error bits are only check once in case there was no error instead of twice (inspired by the NetBSD pcn(4), which additionally predicts the error branch as false). - Fix the debugging output of the RX and TX interrupt handlers; while looping through the descriptors print info about the currently processed one instead of always the previously last used one; remove pointless printing of info about the RX descriptor bits after their values were reset. - Create the DMA tags used to allocate the memory for the init block, descriptors and packet buffers with the alignment the respective NIC actually requires rather than using PAGE_SIZE unconditionally. This might as well fix the alignment of the memory as it seems we do not inherit the alignment constraint from the parent DMA tag. - For the PCI variants double the number of RX descriptors and buffers from 8 to 16 as this minimizes the number of RX overflows im seeing with one NIC-mainboard combination. Nevertheless move reporting of overflows under debugging as they seem unavoidable with some crappy hardware. - Set the software style of the PCI variants to ILACC rather than PCnet-PCI as the former is was am79000.c actually implements. Should not make a difference for this driver though. - Fix the driver name part in the MODULE_DEPEND of the PCI front-end for ether. - Use different device descriptions for PCnet-Home and PCnet-PCI. - Fix some 0/NULL confusion in lance_get(). - Use bus_addr_t for sc_addr and bus_size_t for sc_memsize as these are more appropriate than u_long for these. - Remove the unused LE_DRIVER_NAME macro. - Add a comment describing why we are taking the LE_HTOLE* etc approach instead of using byteorder(9) functions directly. - Improve some comments and fix some wording. MFC after: 2 weeks
2006-05-16 21:04:01 +00:00
if (totlen <= ETHER_HDR_LEN || totlen > LEBLEN - ETHER_CRC_LEN) {
#ifdef LEDEBUG
if_printf(ifp, "invalid packet size %d; dropping\n", totlen);
#endif
return (NULL);
}
MGETHDR(m0, M_DONTWAIT, MT_DATA);
- Revert if_le_pci.c rev. 1.2; although lnc(4) is now gone, le_pci_probe() still should return BUS_PROBE_LOW_PRIORITY instead of BUS_PROBE_DEFAULT in order to give pcn(4) a chance to attach in case it probes after le(4). - Rearrange the code related to RX interrupt handling so that ownership of RX descriptors is immediately returned to the NIC after we have copied the data of the hardware, allowing the NIC to already reuse the descriptor while we are processing the data in ifp->if_input(). This results in a small but measurable increase in RX throughput. As a side-effect, this moves the workaround for the LANCE revision C bug to am7900.c (still off by default as I doubt we will actually encounter such an old chip in a machine running FreeBSD) and the workaround for the bug in the VMware PCnet-PCI emulation to am79000.c, which is now also only compiled on i386 (resulting in a small increase in RX throughput on the other platforms). - Change the RX interrupt handlers so that the descriptor error bits are only check once in case there was no error instead of twice (inspired by the NetBSD pcn(4), which additionally predicts the error branch as false). - Fix the debugging output of the RX and TX interrupt handlers; while looping through the descriptors print info about the currently processed one instead of always the previously last used one; remove pointless printing of info about the RX descriptor bits after their values were reset. - Create the DMA tags used to allocate the memory for the init block, descriptors and packet buffers with the alignment the respective NIC actually requires rather than using PAGE_SIZE unconditionally. This might as well fix the alignment of the memory as it seems we do not inherit the alignment constraint from the parent DMA tag. - For the PCI variants double the number of RX descriptors and buffers from 8 to 16 as this minimizes the number of RX overflows im seeing with one NIC-mainboard combination. Nevertheless move reporting of overflows under debugging as they seem unavoidable with some crappy hardware. - Set the software style of the PCI variants to ILACC rather than PCnet-PCI as the former is was am79000.c actually implements. Should not make a difference for this driver though. - Fix the driver name part in the MODULE_DEPEND of the PCI front-end for ether. - Use different device descriptions for PCnet-Home and PCnet-PCI. - Fix some 0/NULL confusion in lance_get(). - Use bus_addr_t for sc_addr and bus_size_t for sc_memsize as these are more appropriate than u_long for these. - Remove the unused LE_DRIVER_NAME macro. - Add a comment describing why we are taking the LE_HTOLE* etc approach instead of using byteorder(9) functions directly. - Improve some comments and fix some wording. MFC after: 2 weeks
2006-05-16 21:04:01 +00:00
if (m0 == NULL)
return (NULL);
m0->m_pkthdr.rcvif = ifp;
m0->m_pkthdr.len = totlen;
len = MHLEN;
m = m0;
while (totlen > 0) {
if (totlen >= MINCLSIZE) {
MCLGET(m, M_DONTWAIT);
if ((m->m_flags & M_EXT) == 0)
goto bad;
len = MCLBYTES;
}
if (m == m0) {
newdata = (caddr_t)
ALIGN(m->m_data + ETHER_HDR_LEN) - ETHER_HDR_LEN;
len -= newdata - m->m_data;
m->m_data = newdata;
}
m->m_len = len = min(totlen, len);
(*sc->sc_copyfrombuf)(sc, mtod(m, caddr_t), boff, len);
boff += len;
totlen -= len;
if (totlen > 0) {
MGET(newm, M_DONTWAIT, MT_DATA);
if (newm == 0)
goto bad;
len = MLEN;
m = m->m_next = newm;
}
}
return (m0);
bad:
m_freem(m0);
- Revert if_le_pci.c rev. 1.2; although lnc(4) is now gone, le_pci_probe() still should return BUS_PROBE_LOW_PRIORITY instead of BUS_PROBE_DEFAULT in order to give pcn(4) a chance to attach in case it probes after le(4). - Rearrange the code related to RX interrupt handling so that ownership of RX descriptors is immediately returned to the NIC after we have copied the data of the hardware, allowing the NIC to already reuse the descriptor while we are processing the data in ifp->if_input(). This results in a small but measurable increase in RX throughput. As a side-effect, this moves the workaround for the LANCE revision C bug to am7900.c (still off by default as I doubt we will actually encounter such an old chip in a machine running FreeBSD) and the workaround for the bug in the VMware PCnet-PCI emulation to am79000.c, which is now also only compiled on i386 (resulting in a small increase in RX throughput on the other platforms). - Change the RX interrupt handlers so that the descriptor error bits are only check once in case there was no error instead of twice (inspired by the NetBSD pcn(4), which additionally predicts the error branch as false). - Fix the debugging output of the RX and TX interrupt handlers; while looping through the descriptors print info about the currently processed one instead of always the previously last used one; remove pointless printing of info about the RX descriptor bits after their values were reset. - Create the DMA tags used to allocate the memory for the init block, descriptors and packet buffers with the alignment the respective NIC actually requires rather than using PAGE_SIZE unconditionally. This might as well fix the alignment of the memory as it seems we do not inherit the alignment constraint from the parent DMA tag. - For the PCI variants double the number of RX descriptors and buffers from 8 to 16 as this minimizes the number of RX overflows im seeing with one NIC-mainboard combination. Nevertheless move reporting of overflows under debugging as they seem unavoidable with some crappy hardware. - Set the software style of the PCI variants to ILACC rather than PCnet-PCI as the former is was am79000.c actually implements. Should not make a difference for this driver though. - Fix the driver name part in the MODULE_DEPEND of the PCI front-end for ether. - Use different device descriptions for PCnet-Home and PCnet-PCI. - Fix some 0/NULL confusion in lance_get(). - Use bus_addr_t for sc_addr and bus_size_t for sc_memsize as these are more appropriate than u_long for these. - Remove the unused LE_DRIVER_NAME macro. - Add a comment describing why we are taking the LE_HTOLE* etc approach instead of using byteorder(9) functions directly. - Improve some comments and fix some wording. MFC after: 2 weeks
2006-05-16 21:04:01 +00:00
return (NULL);
}
static void
lance_watchdog(void *xsc)
{
struct lance_softc *sc = (struct lance_softc *)xsc;
struct ifnet *ifp = sc->sc_ifp;
LE_LOCK_ASSERT(sc, MA_OWNED);
if (sc->sc_wdog_timer == 0 || --sc->sc_wdog_timer != 0) {
callout_reset(&sc->sc_wdog_ch, hz, lance_watchdog, sc);
return;
}
if_printf(ifp, "device timeout\n");
++ifp->if_oerrors;
lance_init_locked(sc);
}
static int
lance_mediachange(struct ifnet *ifp)
{
struct lance_softc *sc = ifp->if_softc;
if (sc->sc_mediachange) {
/*
* For setting the port in LE_CSR15 the PCnet chips must
* be powered down or stopped and unlike documented may
* not take effect without an initialization. So don't
* invoke (*sc_mediachange) directly here but go through
* lance_init_locked().
*/
LE_LOCK(sc);
lance_stop(sc);
lance_init_locked(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
(*sc->sc_start_locked)(sc);
LE_UNLOCK(sc);
}
return (0);
}
static void
lance_mediastatus(struct ifnet *ifp, struct ifmediareq *ifmr)
{
struct lance_softc *sc = ifp->if_softc;
LE_LOCK(sc);
if (!(ifp->if_flags & IFF_UP)) {
LE_UNLOCK(sc);
return;
}
ifmr->ifm_status = IFM_AVALID;
if (sc->sc_flags & LE_CARRIER)
ifmr->ifm_status |= IFM_ACTIVE;
if (sc->sc_mediastatus)
(*sc->sc_mediastatus)(sc, ifmr);
LE_UNLOCK(sc);
}
/*
* Process an ioctl request.
*/
static int
lance_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
{
struct lance_softc *sc = ifp->if_softc;
struct ifreq *ifr = (struct ifreq *)data;
int error = 0;
switch (cmd) {
case SIOCSIFFLAGS:
LE_LOCK(sc);
if (ifp->if_flags & IFF_PROMISC) {
if (!(sc->sc_flags & LE_PROMISC)) {
sc->sc_flags |= LE_PROMISC;
lance_init_locked(sc);
}
} else if (sc->sc_flags & LE_PROMISC) {
sc->sc_flags &= ~LE_PROMISC;
lance_init_locked(sc);
}
if ((ifp->if_flags & IFF_ALLMULTI) &&
!(sc->sc_flags & LE_ALLMULTI)) {
sc->sc_flags |= LE_ALLMULTI;
lance_init_locked(sc);
} else if (!(ifp->if_flags & IFF_ALLMULTI) &&
(sc->sc_flags & LE_ALLMULTI)) {
sc->sc_flags &= ~LE_ALLMULTI;
lance_init_locked(sc);
}
if (!(ifp->if_flags & IFF_UP) &&
ifp->if_drv_flags & IFF_DRV_RUNNING) {
/*
* If interface is marked down and it is running, then
* stop it.
*/
lance_stop(sc);
} else if (ifp->if_flags & IFF_UP &&
!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
/*
* If interface is marked up and it is stopped, then
* start it.
*/
lance_init_locked(sc);
}
#ifdef LEDEBUG
if (ifp->if_flags & IFF_DEBUG)
sc->sc_flags |= LE_DEBUG;
else
sc->sc_flags &= ~LE_DEBUG;
#endif
LE_UNLOCK(sc);
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
/*
* Multicast list has changed; set the hardware filter
* accordingly.
*/
LE_LOCK(sc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
lance_init_locked(sc);
LE_UNLOCK(sc);
break;
case SIOCGIFMEDIA:
case SIOCSIFMEDIA:
error = ifmedia_ioctl(ifp, ifr, &sc->sc_media, cmd);
break;
default:
error = ether_ioctl(ifp, cmd, data);
break;
}
return (error);
}
/*
* Set up the logical address filter.
*/
void
lance_setladrf(struct lance_softc *sc, uint16_t *af)
{
struct ifnet *ifp = sc->sc_ifp;
struct ifmultiaddr *ifma;
uint32_t crc;
/*
* Set up multicast address filter by passing all multicast addresses
* through a crc generator, and then using the high order 6 bits as an
* index into the 64 bit logical address filter. The high order bit
* selects the word, while the rest of the bits select the bit within
* the word.
*/
if (ifp->if_flags & IFF_PROMISC || sc->sc_flags & LE_ALLMULTI) {
af[0] = af[1] = af[2] = af[3] = 0xffff;
return;
}
af[0] = af[1] = af[2] = af[3] = 0x0000;
if_maddr_rlock(ifp);
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
crc = ether_crc32_le(LLADDR((struct sockaddr_dl *)
ifma->ifma_addr), ETHER_ADDR_LEN);
/* Just want the 6 most significant bits. */
crc >>= 26;
/* Set the corresponding bit in the filter. */
- Revert if_le_pci.c rev. 1.2; although lnc(4) is now gone, le_pci_probe() still should return BUS_PROBE_LOW_PRIORITY instead of BUS_PROBE_DEFAULT in order to give pcn(4) a chance to attach in case it probes after le(4). - Rearrange the code related to RX interrupt handling so that ownership of RX descriptors is immediately returned to the NIC after we have copied the data of the hardware, allowing the NIC to already reuse the descriptor while we are processing the data in ifp->if_input(). This results in a small but measurable increase in RX throughput. As a side-effect, this moves the workaround for the LANCE revision C bug to am7900.c (still off by default as I doubt we will actually encounter such an old chip in a machine running FreeBSD) and the workaround for the bug in the VMware PCnet-PCI emulation to am79000.c, which is now also only compiled on i386 (resulting in a small increase in RX throughput on the other platforms). - Change the RX interrupt handlers so that the descriptor error bits are only check once in case there was no error instead of twice (inspired by the NetBSD pcn(4), which additionally predicts the error branch as false). - Fix the debugging output of the RX and TX interrupt handlers; while looping through the descriptors print info about the currently processed one instead of always the previously last used one; remove pointless printing of info about the RX descriptor bits after their values were reset. - Create the DMA tags used to allocate the memory for the init block, descriptors and packet buffers with the alignment the respective NIC actually requires rather than using PAGE_SIZE unconditionally. This might as well fix the alignment of the memory as it seems we do not inherit the alignment constraint from the parent DMA tag. - For the PCI variants double the number of RX descriptors and buffers from 8 to 16 as this minimizes the number of RX overflows im seeing with one NIC-mainboard combination. Nevertheless move reporting of overflows under debugging as they seem unavoidable with some crappy hardware. - Set the software style of the PCI variants to ILACC rather than PCnet-PCI as the former is was am79000.c actually implements. Should not make a difference for this driver though. - Fix the driver name part in the MODULE_DEPEND of the PCI front-end for ether. - Use different device descriptions for PCnet-Home and PCnet-PCI. - Fix some 0/NULL confusion in lance_get(). - Use bus_addr_t for sc_addr and bus_size_t for sc_memsize as these are more appropriate than u_long for these. - Remove the unused LE_DRIVER_NAME macro. - Add a comment describing why we are taking the LE_HTOLE* etc approach instead of using byteorder(9) functions directly. - Improve some comments and fix some wording. MFC after: 2 weeks
2006-05-16 21:04:01 +00:00
af[crc >> 4] |= LE_HTOLE16(1 << (crc & 0xf));
}
if_maddr_runlock(ifp);
}
/*
* Routines for accessing the transmit and receive buffers.
* The various CPU and adapter configurations supported by this
* driver require three different access methods for buffers
* and descriptors:
* (1) contig (contiguous data; no padding),
* (2) gap2 (two bytes of data followed by two bytes of padding),
* (3) gap16 (16 bytes of data followed by 16 bytes of padding).
*/
/*
* contig: contiguous data with no padding.
*
* Buffers may have any alignment.
*/
void
lance_copytobuf_contig(struct lance_softc *sc, void *from, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
/*
* Just call memcpy() to do the work.
*/
memcpy(buf + boff, from, len);
}
void
lance_copyfrombuf_contig(struct lance_softc *sc, void *to, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
/*
* Just call memcpy() to do the work.
*/
memcpy(to, buf + boff, len);
}
void
lance_zerobuf_contig(struct lance_softc *sc, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
/*
* Just let memset() do the work
*/
memset(buf + boff, 0, len);
}
#if 0
/*
* Examples only; duplicate these and tweak (if necessary) in
* machine-specific front-ends.
*/
/*
* gap2: two bytes of data followed by two bytes of pad.
*
* Buffers must be 4-byte aligned. The code doesn't worry about
* doing an extra byte.
*/
static void
lance_copytobuf_gap2(struct lance_softc *sc, void *fromv, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
caddr_t from = fromv;
volatile uint16_t *bptr;
if (boff & 0x1) {
/* Handle unaligned first byte. */
bptr = ((volatile uint16_t *)buf) + (boff - 1);
*bptr = (*from++ << 8) | (*bptr & 0xff);
bptr += 2;
len--;
} else
bptr = ((volatile uint16_t *)buf) + boff;
while (len > 1) {
*bptr = (from[1] << 8) | (from[0] & 0xff);
bptr += 2;
from += 2;
len -= 2;
}
if (len == 1)
*bptr = (uint16_t)*from;
}
static void
lance_copyfrombuf_gap2(struct lance_softc *sc, void *tov, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
caddr_t to = tov;
volatile uint16_t *bptr;
uint16_t tmp;
if (boff & 0x1) {
/* Handle unaligned first byte. */
bptr = ((volatile uint16_t *)buf) + (boff - 1);
*to++ = (*bptr >> 8) & 0xff;
bptr += 2;
len--;
} else
bptr = ((volatile uint16_t *)buf) + boff;
while (len > 1) {
tmp = *bptr;
*to++ = tmp & 0xff;
*to++ = (tmp >> 8) & 0xff;
bptr += 2;
len -= 2;
}
if (len == 1)
*to = *bptr & 0xff;
}
static void
lance_zerobuf_gap2(struct lance_softc *sc, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
volatile uint16_t *bptr;
if ((unsigned)boff & 0x1) {
bptr = ((volatile uint16_t *)buf) + (boff - 1);
*bptr &= 0xff;
bptr += 2;
len--;
} else
bptr = ((volatile uint16_t *)buf) + boff;
while (len > 0) {
*bptr = 0;
bptr += 2;
len -= 2;
}
}
/*
* gap16: 16 bytes of data followed by 16 bytes of pad.
*
* Buffers must be 32-byte aligned.
*/
static void
lance_copytobuf_gap16(struct lance_softc *sc, void *fromv, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
caddr_t bptr, from = fromv;
int xfer;
bptr = buf + ((boff << 1) & ~0x1f);
boff &= 0xf;
xfer = min(len, 16 - boff);
while (len > 0) {
memcpy(bptr + boff, from, xfer);
from += xfer;
bptr += 32;
boff = 0;
len -= xfer;
xfer = min(len, 16);
}
}
static void
lance_copyfrombuf_gap16(struct lance_softc *sc, void *tov, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
caddr_t bptr, to = tov;
int xfer;
bptr = buf + ((boff << 1) & ~0x1f);
boff &= 0xf;
xfer = min(len, 16 - boff);
while (len > 0) {
memcpy(to, bptr + boff, xfer);
to += xfer;
bptr += 32;
boff = 0;
len -= xfer;
xfer = min(len, 16);
}
}
static void
lance_zerobuf_gap16(struct lance_softc *sc, int boff, int len)
{
volatile caddr_t buf = sc->sc_mem;
caddr_t bptr;
int xfer;
bptr = buf + ((boff << 1) & ~0x1f);
boff &= 0xf;
xfer = min(len, 16 - boff);
while (len > 0) {
memset(bptr + boff, 0, xfer);
bptr += 32;
boff = 0;
len -= xfer;
xfer = min(len, 16);
}
}
#endif /* Example only */