freebsd-nq/sys/dev/le/am79900.c

657 lines
18 KiB
C
Raw Normal View History

/* $NetBSD: am79900.c,v 1.17 2005/12/24 20:27:29 perry Exp $ */
/*-
* Copyright (c) 1997 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Jason R. Thorpe.
*
* 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.
*
* 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
*/
/*-
* Copyright (c) 1998
* Matthias Drochner. All rights reserved.
* Copyright (c) 1995 Charles M. Hannum. 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. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. 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/mbuf.h>
#include <sys/mutex.h>
#include <sys/socket.h>
#include <net/bpf.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_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>
#include <dev/le/am79900reg.h>
#include <dev/le/am79900var.h>
static void am79900_meminit(struct lance_softc *);
static void am79900_rint(struct lance_softc *);
static void am79900_tint(struct lance_softc *);
static void am79900_start_locked(struct lance_softc *sc);
#ifdef LEDEBUG
static void am79900_recv_print(struct lance_softc *, int);
static void am79900_xmit_print(struct lance_softc *, int);
#endif
int
am79900_config(struct am79900_softc *sc, const char* name, int unit)
{
int error, mem;
sc->lsc.sc_meminit = am79900_meminit;
sc->lsc.sc_start_locked = am79900_start_locked;
error = lance_config(&sc->lsc, name, unit);
if (error != 0)
return (error);
mem = 0;
sc->lsc.sc_initaddr = mem;
mem += sizeof(struct leinit);
sc->lsc.sc_rmdaddr = mem;
mem += sizeof(struct lermd) * sc->lsc.sc_nrbuf;
sc->lsc.sc_tmdaddr = mem;
mem += sizeof(struct letmd) * sc->lsc.sc_ntbuf;
sc->lsc.sc_rbufaddr = mem;
mem += LEBLEN * sc->lsc.sc_nrbuf;
sc->lsc.sc_tbufaddr = mem;
mem += LEBLEN * sc->lsc.sc_ntbuf;
if (mem > sc->lsc.sc_memsize)
panic("%s: memsize", __func__);
lance_attach(&sc->lsc);
return (0);
}
void
am79900_detach(struct am79900_softc *sc)
{
lance_detach(&sc->lsc);
}
/*
* Set up the initialization block and the descriptor rings.
*/
static void
am79900_meminit(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
struct leinit init;
struct lermd rmd;
struct letmd tmd;
u_long a;
int bix;
LE_LOCK_ASSERT(sc, MA_OWNED);
if (ifp->if_flags & IFF_PROMISC)
init.init_mode = LE_HTOLE32(LE_MODE_NORMAL | LE_MODE_PROM);
else
init.init_mode = LE_HTOLE32(LE_MODE_NORMAL);
init.init_mode |= LE_HTOLE32(((ffs(sc->sc_ntbuf) - 1) << 28) |
((ffs(sc->sc_nrbuf) - 1) << 20));
init.init_padr[0] = LE_HTOLE32(sc->sc_enaddr[0] |
(sc->sc_enaddr[1] << 8) | (sc->sc_enaddr[2] << 16) |
(sc->sc_enaddr[3] << 24));
init.init_padr[1] = LE_HTOLE32(sc->sc_enaddr[4] |
(sc->sc_enaddr[5] << 8));
lance_setladrf(sc, init.init_ladrf);
sc->sc_last_rd = 0;
sc->sc_first_td = sc->sc_last_td = sc->sc_no_td = 0;
a = sc->sc_addr + LE_RMDADDR(sc, 0);
init.init_rdra = LE_HTOLE32(a);
a = sc->sc_addr + LE_TMDADDR(sc, 0);
init.init_tdra = LE_HTOLE32(a);
(*sc->sc_copytodesc)(sc, &init, LE_INITADDR(sc), sizeof(init));
/*
* Set up receive ring descriptors.
*/
for (bix = 0; bix < sc->sc_nrbuf; bix++) {
a = sc->sc_addr + LE_RBUFADDR(sc, bix);
rmd.rmd0 = LE_HTOLE32(a);
rmd.rmd1 = LE_HTOLE32(LE_R1_OWN | LE_R1_ONES |
(-LEBLEN & 0xfff));
rmd.rmd2 = 0;
rmd.rmd3 = 0;
(*sc->sc_copytodesc)(sc, &rmd, LE_RMDADDR(sc, bix),
sizeof(rmd));
}
/*
* Set up transmit ring descriptors.
*/
for (bix = 0; bix < sc->sc_ntbuf; bix++) {
a = sc->sc_addr + LE_TBUFADDR(sc, bix);
tmd.tmd0 = LE_HTOLE32(a);
tmd.tmd1 = LE_HTOLE32(LE_T1_ONES);
tmd.tmd2 = 0;
tmd.tmd3 = 0;
(*sc->sc_copytodesc)(sc, &tmd, LE_TMDADDR(sc, bix),
sizeof(tmd));
}
}
static inline void
am79900_rint(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
- 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 *m;
struct lermd rmd;
uint32_t rmd1;
int bix, rp;
- 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 defined(__i386__) && !defined(PC98)
struct ether_header *eh;
#endif
bix = sc->sc_last_rd;
/* Process all buffers with valid data. */
for (;;) {
rp = LE_RMDADDR(sc, bix);
(*sc->sc_copyfromdesc)(sc, &rmd, rp, sizeof(rmd));
rmd1 = LE_LE32TOH(rmd.rmd1);
if (rmd1 & LE_R1_OWN)
break;
- 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
m = NULL;
if ((rmd1 & (LE_R1_ERR | LE_R1_STP | LE_R1_ENP)) !=
(LE_R1_STP | LE_R1_ENP)){
if (rmd1 & LE_R1_ERR) {
#ifdef LEDEBUG
- 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 (rmd1 & LE_R1_ENP) {
if ((rmd1 & LE_R1_OFLO) == 0) {
if (rmd1 & LE_R1_FRAM)
if_printf(ifp,
"framing error\n");
if (rmd1 & LE_R1_CRC)
if_printf(ifp,
"crc mismatch\n");
}
} else
if (rmd1 & LE_R1_OFLO)
if_printf(ifp, "overflow\n");
#endif
- 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 (rmd1 & LE_R1_BUFF)
if_printf(ifp,
"receive buffer error\n");
} else if ((rmd1 & (LE_R1_STP | LE_R1_ENP)) !=
(LE_R1_STP | LE_R1_ENP))
if_printf(ifp, "dropping chained buffer\n");
} else {
#ifdef LEDEBUG
if (sc->sc_flags & LE_DEBUG)
- 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
am79900_recv_print(sc, bix);
#endif
- 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
/* Pull the packet off the interface. */
m = lance_get(sc, LE_RBUFADDR(sc, bix),
(LE_LE32TOH(rmd.rmd2) & 0xfff) - ETHER_CRC_LEN);
}
rmd.rmd1 = LE_HTOLE32(LE_R1_OWN | LE_R1_ONES |
(-LEBLEN & 0xfff));
rmd.rmd2 = 0;
rmd.rmd3 = 0;
(*sc->sc_copytodesc)(sc, &rmd, rp, sizeof(rmd));
if (++bix == sc->sc_nrbuf)
bix = 0;
- 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 (m != NULL) {
ifp->if_ipackets++;
#if defined(__i386__) && !defined(PC98)
/*
* The VMware LANCE does not present IFF_SIMPLEX
* behavior on multicast packets. Thus drop the
* packet if it is from ourselves.
*/
eh = mtod(m, struct ether_header *);
if (!ether_cmp(eh->ether_shost, sc->sc_enaddr)) {
m_freem(m);
continue;
}
#endif
/* Pass the packet up. */
LE_UNLOCK(sc);
(*ifp->if_input)(ifp, m);
LE_LOCK(sc);
} else
ifp->if_ierrors++;
}
sc->sc_last_rd = bix;
}
static inline void
am79900_tint(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
struct letmd tmd;
uint32_t tmd1, tmd2;
int bix;
bix = sc->sc_first_td;
for (;;) {
if (sc->sc_no_td <= 0)
break;
(*sc->sc_copyfromdesc)(sc, &tmd, LE_TMDADDR(sc, bix),
sizeof(tmd));
- 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
tmd1 = LE_LE32TOH(tmd.tmd1);
#ifdef LEDEBUG
if (sc->sc_flags & LE_DEBUG)
if_printf(ifp, "trans tmd: "
"adr %08x, flags/blen %08x\n",
- 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
LE_LE32TOH(tmd.tmd0), tmd1);
#endif
if (tmd1 & LE_T1_OWN)
break;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
if (tmd1 & LE_T1_ERR) {
- 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
tmd2 = LE_LE32TOH(tmd.tmd2);
if (tmd2 & LE_T2_BUFF)
if_printf(ifp, "transmit buffer error\n");
else if (tmd2 & LE_T2_UFLO)
if_printf(ifp, "underflow\n");
if (tmd2 & (LE_T2_BUFF | LE_T2_UFLO)) {
lance_init_locked(sc);
return;
}
if (tmd2 & LE_T2_LCAR) {
if (sc->sc_flags & LE_CARRIER)
if_link_state_change(ifp,
LINK_STATE_DOWN);
sc->sc_flags &= ~LE_CARRIER;
if (sc->sc_nocarrier)
(*sc->sc_nocarrier)(sc);
else
if_printf(ifp, "lost carrier\n");
}
if (tmd2 & LE_T2_LCOL)
ifp->if_collisions++;
if (tmd2 & LE_T2_RTRY) {
#ifdef LEDEBUG
if_printf(ifp, "excessive collisions\n");
#endif
ifp->if_collisions += 16;
}
ifp->if_oerrors++;
} else {
if (tmd1 & LE_T1_ONE)
ifp->if_collisions++;
else if (tmd1 & LE_T1_MORE)
/* Real number is unknown. */
ifp->if_collisions += 2;
ifp->if_opackets++;
}
if (++bix == sc->sc_ntbuf)
bix = 0;
--sc->sc_no_td;
}
sc->sc_first_td = bix;
sc->sc_wdog_timer = sc->sc_no_td > 0 ? 5 : 0;
}
/*
* Controller interrupt
*/
void
am79900_intr(void *arg)
{
struct lance_softc *sc = arg;
struct ifnet *ifp = sc->sc_ifp;
uint16_t isr;
LE_LOCK(sc);
if (sc->sc_hwintr && (*sc->sc_hwintr)(sc) == -1) {
ifp->if_ierrors++;
lance_init_locked(sc);
LE_UNLOCK(sc);
return;
}
isr = (*sc->sc_rdcsr)(sc, LE_CSR0);
#if defined(LEDEBUG) && LEDEBUG > 1
if (sc->sc_flags & LE_DEBUG)
if_printf(ifp, "%s: entering with isr=%04x\n", __func__, isr);
#endif
if ((isr & LE_C0_INTR) == 0) {
LE_UNLOCK(sc);
return;
}
/*
* Clear interrupt source flags and turn off interrupts. If we
* don't clear these flags before processing their sources we
- 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
* could completely miss some interrupt events as the NIC can
* change these flags while we're in this handler. We toggle
* the interrupt enable bit in order to keep receiving them
* (some chips work without this, some don't).
*/
(*sc->sc_wrcsr)(sc, LE_CSR0, isr & ~(LE_C0_INEA | LE_C0_TDMD |
LE_C0_STOP | LE_C0_STRT | LE_C0_INIT));
if (isr & LE_C0_ERR) {
if (isr & LE_C0_BABL) {
#ifdef LEDEBUG
if_printf(ifp, "babble\n");
#endif
ifp->if_oerrors++;
}
#if 0
if (isr & LE_C0_CERR) {
if_printf(ifp, "collision error\n");
ifp->if_collisions++;
}
#endif
if (isr & LE_C0_MISS) {
#ifdef LEDEBUG
if_printf(ifp, "missed packet\n");
#endif
ifp->if_ierrors++;
}
if (isr & LE_C0_MERR) {
if_printf(ifp, "memory error\n");
lance_init_locked(sc);
LE_UNLOCK(sc);
return;
}
}
if ((isr & LE_C0_RXON) == 0) {
if_printf(ifp, "receiver disabled\n");
ifp->if_ierrors++;
lance_init_locked(sc);
LE_UNLOCK(sc);
return;
}
if ((isr & LE_C0_TXON) == 0) {
if_printf(ifp, "transmitter disabled\n");
ifp->if_oerrors++;
lance_init_locked(sc);
LE_UNLOCK(sc);
return;
}
/*
* Pretend we have carrier; if we don't this will be cleared shortly.
*/
if (!(sc->sc_flags & LE_CARRIER))
if_link_state_change(ifp, LINK_STATE_UP);
sc->sc_flags |= LE_CARRIER;
if (isr & LE_C0_RINT)
am79900_rint(sc);
if (isr & LE_C0_TINT)
am79900_tint(sc);
/* Enable interrupts again. */
(*sc->sc_wrcsr)(sc, LE_CSR0, LE_C0_INEA);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
am79900_start_locked(sc);
LE_UNLOCK(sc);
}
/*
* Set up output on interface.
* Get another datagram to send off of the interface queue, and map it to the
* interface before starting the output.
*/
static void
am79900_start_locked(struct lance_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
struct letmd tmd;
struct mbuf *m;
int bix, enq, len, rp;
LE_LOCK_ASSERT(sc, MA_OWNED);
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING)
return;
bix = sc->sc_last_td;
enq = 0;
for (; sc->sc_no_td < sc->sc_ntbuf &&
!IFQ_DRV_IS_EMPTY(&ifp->if_snd);) {
rp = LE_TMDADDR(sc, bix);
(*sc->sc_copyfromdesc)(sc, &tmd, rp, sizeof(tmd));
if (LE_LE32TOH(tmd.tmd1) & LE_T1_OWN) {
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
if_printf(ifp,
"missing buffer, no_td = %d, last_td = %d\n",
sc->sc_no_td, sc->sc_last_td);
}
IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
if (m == 0)
break;
/*
* If BPF is listening on this interface, let it see the packet
* before we commit it to the wire.
*/
BPF_MTAP(ifp, m);
/*
* Copy the mbuf chain into the transmit buffer.
*/
len = lance_put(sc, LE_TBUFADDR(sc, bix), m);
#ifdef LEDEBUG
if (len > ETHERMTU + ETHER_HDR_LEN)
if_printf(ifp, "packet length %d\n", len);
#endif
/*
* Init transmit registers, and set transmit start flag.
*/
tmd.tmd1 = LE_HTOLE32(LE_T1_OWN | LE_T1_STP | LE_T1_ENP |
LE_T1_ONES | (-len & 0xfff));
tmd.tmd2 = 0;
tmd.tmd3 = 0;
(*sc->sc_copytodesc)(sc, &tmd, rp, sizeof(tmd));
#ifdef LEDEBUG
if (sc->sc_flags & LE_DEBUG)
- 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
am79900_xmit_print(sc, bix);
#endif
(*sc->sc_wrcsr)(sc, LE_CSR0, LE_C0_INEA | LE_C0_TDMD);
enq++;
if (++bix == sc->sc_ntbuf)
bix = 0;
if (++sc->sc_no_td == sc->sc_ntbuf) {
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
break;
}
}
sc->sc_last_td = bix;
if (enq > 0)
sc->sc_wdog_timer = 5;
}
#ifdef LEDEBUG
static void
am79900_recv_print(struct lance_softc *sc, int no)
{
struct ifnet *ifp = sc->sc_ifp;
struct ether_header eh;
struct lermd rmd;
uint16_t len;
(*sc->sc_copyfromdesc)(sc, &rmd, LE_RMDADDR(sc, no), sizeof(rmd));
len = LE_LE32TOH(rmd.rmd2) & 0xfff;
if_printf(ifp, "receive buffer %d, len = %d\n", no, len);
if_printf(ifp, "status %04x\n", (*sc->sc_rdcsr)(sc, LE_CSR0));
if_printf(ifp, "adr %08x, flags/blen %08x\n", LE_LE32TOH(rmd.rmd0),
LE_LE32TOH(rmd.rmd1));
if (len - ETHER_CRC_LEN >= sizeof(eh)) {
(*sc->sc_copyfrombuf)(sc, &eh, LE_RBUFADDR(sc, no), sizeof(eh));
if_printf(ifp, "dst %s", ether_sprintf(eh.ether_dhost));
printf(" src %s type %04x\n", ether_sprintf(eh.ether_shost),
ntohs(eh.ether_type));
}
}
static void
am79900_xmit_print(struct lance_softc *sc, int no)
{
struct ifnet *ifp = sc->sc_ifp;
struct ether_header eh;
struct letmd tmd;
uint16_t len;
(*sc->sc_copyfromdesc)(sc, &tmd, LE_TMDADDR(sc, no), sizeof(tmd));
len = -(LE_LE32TOH(tmd.tmd1) & 0xfff);
if_printf(ifp, "transmit buffer %d, len = %d\n", no, len);
if_printf(ifp, "status %04x\n", (*sc->sc_rdcsr)(sc, LE_CSR0));
if_printf(ifp, "adr %08x, flags/blen %08x\n", LE_LE32TOH(tmd.tmd0),
LE_LE32TOH(tmd.tmd1));
if (len >= sizeof(eh)) {
(*sc->sc_copyfrombuf)(sc, &eh, LE_TBUFADDR(sc, no), sizeof(eh));
if_printf(ifp, "dst %s", ether_sprintf(eh.ether_dhost));
printf(" src %s type %04x\n", ether_sprintf(eh.ether_shost),
ntohs(eh.ether_type));
}
}
#endif /* LEDEBUG */