freebsd-nq/sys/pci/if_ste.c

1750 lines
38 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1997, 1998, 1999
* Bill Paul <wpaul@ctr.columbia.edu>. All rights reserved.
*
* 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 Bill Paul.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY Bill Paul 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 Bill Paul OR THE VOICES IN HIS HEAD
* 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.
*/
2003-04-03 21:36:33 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
#endif
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sockio.h>
#include <sys/mbuf.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
2004-05-30 20:00:41 +00:00
#include <sys/module.h>
#include <sys/socket.h>
#include <sys/sysctl.h>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/ethernet.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
#include <net/if_vlan_var.h>
#include <net/bpf.h>
#include <vm/vm.h> /* for vtophys */
#include <vm/pmap.h> /* for vtophys */
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
#include <dev/mii/mii.h>
#include <dev/mii/miivar.h>
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
/* "device miibus" required. See GENERIC if you get errors here. */
#include "miibus_if.h"
#define STE_USEIOSPACE
#include <pci/if_stereg.h>
MODULE_DEPEND(ste, pci, 1, 1, 1);
MODULE_DEPEND(ste, ether, 1, 1, 1);
MODULE_DEPEND(ste, miibus, 1, 1, 1);
/*
* Various supported device vendors/types and their names.
*/
static struct ste_type ste_devs[] = {
{ ST_VENDORID, ST_DEVICEID_ST201, "Sundance ST201 10/100BaseTX" },
{ DL_VENDORID, DL_DEVICEID_DL10050, "D-Link DL10050 10/100BaseTX" },
{ 0, 0, NULL }
};
static int ste_probe(device_t);
static int ste_attach(device_t);
static int ste_detach(device_t);
static void ste_init(void *);
static void ste_init_locked(struct ste_softc *);
static void ste_intr(void *);
static void ste_rxeoc(struct ste_softc *);
static void ste_rxeof(struct ste_softc *);
static void ste_txeoc(struct ste_softc *);
static void ste_txeof(struct ste_softc *);
static void ste_stats_update(void *);
static void ste_stop(struct ste_softc *);
static void ste_reset(struct ste_softc *);
static int ste_ioctl(struct ifnet *, u_long, caddr_t);
static int ste_encap(struct ste_softc *, struct ste_chain *, struct mbuf *);
static void ste_start(struct ifnet *);
static void ste_start_locked(struct ifnet *);
static void ste_watchdog(struct ifnet *);
static void ste_shutdown(device_t);
static int ste_newbuf(struct ste_softc *, struct ste_chain_onefrag *,
struct mbuf *);
static int ste_ifmedia_upd(struct ifnet *);
static void ste_ifmedia_upd_locked(struct ifnet *);
static void ste_ifmedia_sts(struct ifnet *, struct ifmediareq *);
static void ste_mii_sync(struct ste_softc *);
static void ste_mii_send(struct ste_softc *, u_int32_t, int);
static int ste_mii_readreg(struct ste_softc *, struct ste_mii_frame *);
static int ste_mii_writereg(struct ste_softc *, struct ste_mii_frame *);
static int ste_miibus_readreg(device_t, int, int);
static int ste_miibus_writereg(device_t, int, int, int);
static void ste_miibus_statchg(device_t);
static int ste_eeprom_wait(struct ste_softc *);
static int ste_read_eeprom(struct ste_softc *, caddr_t, int, int, int);
static void ste_wait(struct ste_softc *);
static void ste_setmulti(struct ste_softc *);
static int ste_init_rx_list(struct ste_softc *);
static void ste_init_tx_list(struct ste_softc *);
#ifdef STE_USEIOSPACE
#define STE_RES SYS_RES_IOPORT
#define STE_RID STE_PCI_LOIO
#else
#define STE_RES SYS_RES_MEMORY
#define STE_RID STE_PCI_LOMEM
#endif
static device_method_t ste_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, ste_probe),
DEVMETHOD(device_attach, ste_attach),
DEVMETHOD(device_detach, ste_detach),
DEVMETHOD(device_shutdown, ste_shutdown),
/* bus interface */
DEVMETHOD(bus_print_child, bus_generic_print_child),
DEVMETHOD(bus_driver_added, bus_generic_driver_added),
/* MII interface */
DEVMETHOD(miibus_readreg, ste_miibus_readreg),
DEVMETHOD(miibus_writereg, ste_miibus_writereg),
DEVMETHOD(miibus_statchg, ste_miibus_statchg),
{ 0, 0 }
};
static driver_t ste_driver = {
"ste",
ste_methods,
sizeof(struct ste_softc)
};
static devclass_t ste_devclass;
DRIVER_MODULE(ste, pci, ste_driver, ste_devclass, 0, 0);
Un-do the changes to the DRIVER_MODULE() declarations in these drivers. This whole idea isn't going to work until somebody makes the bus/kld code smarter. The idea here is to change the module's internal name from "foo" to "if_foo" so that ifconfig can tell a network driver from a non-network one. However doing this doesn't work correctly no matter how you slice it. For everything to work, you have to change the name in both the driver_t struct and the DRIVER_MODULE() declaration. The problems are: - If you change the name in both places, then the kernel thinks that the device's name is now "if_foo", so you get things like: if_foo0: <FOO ethernet> irq foo at device foo on pcifoo if_foo0: Ethernet address: foo:foo:foo:foo:foo:foo This is bogus. Now the device name doesn't agree with the logical interface name. There's no reason for this, and it violates the principle of least astonishment. - If you leave the name in the driver_t struct as "foo" and only change the names in the DRIVER_MODULE() declaration to "if_foo" then attaching drivers to child devices doesn't work because the names don't agree. This breaks miibus: drivers that need to have miibuses and PHY drivers attached never get them. In other words: damned if you do, damned if you don't. This needs to be thought through some more. Since the drivers that use miibus are broken, I have to change these all back in order to make them work again. Yes this will stop ifconfig from being able to demand load driver modules. On the whole, I'd rather have that than having the drivers not work at all.
1999-09-20 19:06:45 +00:00
DRIVER_MODULE(miibus, ste, miibus_driver, miibus_devclass, 0, 0);
SYSCTL_NODE(_hw, OID_AUTO, ste, CTLFLAG_RD, 0, "if_ste parameters");
static int ste_rxsyncs;
SYSCTL_INT(_hw_ste, OID_AUTO, rxsyncs, CTLFLAG_RW, &ste_rxsyncs, 0, "");
#define STE_SETBIT4(sc, reg, x) \
CSR_WRITE_4(sc, reg, CSR_READ_4(sc, reg) | (x))
#define STE_CLRBIT4(sc, reg, x) \
CSR_WRITE_4(sc, reg, CSR_READ_4(sc, reg) & ~(x))
#define STE_SETBIT2(sc, reg, x) \
CSR_WRITE_2(sc, reg, CSR_READ_2(sc, reg) | (x))
#define STE_CLRBIT2(sc, reg, x) \
CSR_WRITE_2(sc, reg, CSR_READ_2(sc, reg) & ~(x))
#define STE_SETBIT1(sc, reg, x) \
CSR_WRITE_1(sc, reg, CSR_READ_1(sc, reg) | (x))
#define STE_CLRBIT1(sc, reg, x) \
CSR_WRITE_1(sc, reg, CSR_READ_1(sc, reg) & ~(x))
#define MII_SET(x) STE_SETBIT1(sc, STE_PHYCTL, x)
#define MII_CLR(x) STE_CLRBIT1(sc, STE_PHYCTL, x)
/*
* Sync the PHYs by setting data bit and strobing the clock 32 times.
*/
static void
ste_mii_sync(sc)
struct ste_softc *sc;
{
register int i;
MII_SET(STE_PHYCTL_MDIR|STE_PHYCTL_MDATA);
for (i = 0; i < 32; i++) {
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
}
return;
}
/*
* Clock a series of bits through the MII.
*/
static void
ste_mii_send(sc, bits, cnt)
struct ste_softc *sc;
u_int32_t bits;
int cnt;
{
int i;
MII_CLR(STE_PHYCTL_MCLK);
for (i = (0x1 << (cnt - 1)); i; i >>= 1) {
if (bits & i) {
MII_SET(STE_PHYCTL_MDATA);
} else {
MII_CLR(STE_PHYCTL_MDATA);
}
DELAY(1);
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
MII_SET(STE_PHYCTL_MCLK);
}
}
/*
* Read an PHY register through the MII.
*/
static int
ste_mii_readreg(sc, frame)
struct ste_softc *sc;
struct ste_mii_frame *frame;
{
int i, ack;
/*
* Set up frame for RX.
*/
frame->mii_stdelim = STE_MII_STARTDELIM;
frame->mii_opcode = STE_MII_READOP;
frame->mii_turnaround = 0;
frame->mii_data = 0;
CSR_WRITE_2(sc, STE_PHYCTL, 0);
/*
* Turn on data xmit.
*/
MII_SET(STE_PHYCTL_MDIR);
ste_mii_sync(sc);
/*
* Send command/address info.
*/
ste_mii_send(sc, frame->mii_stdelim, 2);
ste_mii_send(sc, frame->mii_opcode, 2);
ste_mii_send(sc, frame->mii_phyaddr, 5);
ste_mii_send(sc, frame->mii_regaddr, 5);
/* Turn off xmit. */
MII_CLR(STE_PHYCTL_MDIR);
/* Idle bit */
MII_CLR((STE_PHYCTL_MCLK|STE_PHYCTL_MDATA));
DELAY(1);
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
/* Check for ack */
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
ack = CSR_READ_2(sc, STE_PHYCTL) & STE_PHYCTL_MDATA;
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
/*
* Now try reading data bits. If the ack failed, we still
* need to clock through 16 cycles to keep the PHY(s) in sync.
*/
if (ack) {
for(i = 0; i < 16; i++) {
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
}
goto fail;
}
for (i = 0x8000; i; i >>= 1) {
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
if (!ack) {
if (CSR_READ_2(sc, STE_PHYCTL) & STE_PHYCTL_MDATA)
frame->mii_data |= i;
DELAY(1);
}
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
}
fail:
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
if (ack)
return(1);
return(0);
}
/*
* Write to a PHY register through the MII.
*/
static int
ste_mii_writereg(sc, frame)
struct ste_softc *sc;
struct ste_mii_frame *frame;
{
/*
* Set up frame for TX.
*/
frame->mii_stdelim = STE_MII_STARTDELIM;
frame->mii_opcode = STE_MII_WRITEOP;
frame->mii_turnaround = STE_MII_TURNAROUND;
/*
* Turn on data output.
*/
MII_SET(STE_PHYCTL_MDIR);
ste_mii_sync(sc);
ste_mii_send(sc, frame->mii_stdelim, 2);
ste_mii_send(sc, frame->mii_opcode, 2);
ste_mii_send(sc, frame->mii_phyaddr, 5);
ste_mii_send(sc, frame->mii_regaddr, 5);
ste_mii_send(sc, frame->mii_turnaround, 2);
ste_mii_send(sc, frame->mii_data, 16);
/* Idle bit. */
MII_SET(STE_PHYCTL_MCLK);
DELAY(1);
MII_CLR(STE_PHYCTL_MCLK);
DELAY(1);
/*
* Turn off xmit.
*/
MII_CLR(STE_PHYCTL_MDIR);
return(0);
}
static int
ste_miibus_readreg(dev, phy, reg)
device_t dev;
int phy, reg;
{
struct ste_softc *sc;
struct ste_mii_frame frame;
sc = device_get_softc(dev);
if ( sc->ste_one_phy && phy != 0 )
return (0);
bzero((char *)&frame, sizeof(frame));
frame.mii_phyaddr = phy;
frame.mii_regaddr = reg;
ste_mii_readreg(sc, &frame);
return(frame.mii_data);
}
static int
ste_miibus_writereg(dev, phy, reg, data)
device_t dev;
int phy, reg, data;
{
struct ste_softc *sc;
struct ste_mii_frame frame;
sc = device_get_softc(dev);
bzero((char *)&frame, sizeof(frame));
frame.mii_phyaddr = phy;
frame.mii_regaddr = reg;
frame.mii_data = data;
ste_mii_writereg(sc, &frame);
return(0);
}
static void
ste_miibus_statchg(dev)
device_t dev;
{
struct ste_softc *sc;
struct mii_data *mii;
sc = device_get_softc(dev);
mii = device_get_softc(sc->ste_miibus);
if ((mii->mii_media_active & IFM_GMASK) == IFM_FDX) {
STE_SETBIT2(sc, STE_MACCTL0, STE_MACCTL0_FULLDUPLEX);
} else {
STE_CLRBIT2(sc, STE_MACCTL0, STE_MACCTL0_FULLDUPLEX);
}
return;
}
static int
ste_ifmedia_upd(ifp)
struct ifnet *ifp;
{
struct ste_softc *sc;
sc = ifp->if_softc;
STE_LOCK(sc);
ste_ifmedia_upd_locked(ifp);
STE_UNLOCK(sc);
return(0);
}
static void
ste_ifmedia_upd_locked(ifp)
struct ifnet *ifp;
{
struct ste_softc *sc;
struct mii_data *mii;
sc = ifp->if_softc;
STE_LOCK_ASSERT(sc);
mii = device_get_softc(sc->ste_miibus);
sc->ste_link = 0;
if (mii->mii_instance) {
struct mii_softc *miisc;
LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
mii_phy_reset(miisc);
}
mii_mediachg(mii);
}
static void
ste_ifmedia_sts(ifp, ifmr)
struct ifnet *ifp;
struct ifmediareq *ifmr;
{
struct ste_softc *sc;
struct mii_data *mii;
sc = ifp->if_softc;
mii = device_get_softc(sc->ste_miibus);
STE_LOCK(sc);
mii_pollstat(mii);
ifmr->ifm_active = mii->mii_media_active;
ifmr->ifm_status = mii->mii_media_status;
STE_UNLOCK(sc);
return;
}
static void
ste_wait(sc)
struct ste_softc *sc;
{
register int i;
for (i = 0; i < STE_TIMEOUT; i++) {
if (!(CSR_READ_4(sc, STE_DMACTL) & STE_DMACTL_DMA_HALTINPROG))
break;
}
if (i == STE_TIMEOUT)
device_printf(sc->ste_dev, "command never completed!\n");
return;
}
/*
* The EEPROM is slow: give it time to come ready after issuing
* it a command.
*/
static int
ste_eeprom_wait(sc)
struct ste_softc *sc;
{
int i;
DELAY(1000);
for (i = 0; i < 100; i++) {
if (CSR_READ_2(sc, STE_EEPROM_CTL) & STE_EECTL_BUSY)
DELAY(1000);
else
break;
}
if (i == 100) {
device_printf(sc->ste_dev, "eeprom failed to come ready\n");
return(1);
}
return(0);
}
/*
* Read a sequence of words from the EEPROM. Note that ethernet address
* data is stored in the EEPROM in network byte order.
*/
static int
ste_read_eeprom(sc, dest, off, cnt, swap)
struct ste_softc *sc;
caddr_t dest;
int off;
int cnt;
int swap;
{
int err = 0, i;
u_int16_t word = 0, *ptr;
if (ste_eeprom_wait(sc))
return(1);
for (i = 0; i < cnt; i++) {
CSR_WRITE_2(sc, STE_EEPROM_CTL, STE_EEOPCODE_READ | (off + i));
err = ste_eeprom_wait(sc);
if (err)
break;
word = CSR_READ_2(sc, STE_EEPROM_DATA);
ptr = (u_int16_t *)(dest + (i * 2));
if (swap)
*ptr = ntohs(word);
else
*ptr = word;
}
return(err ? 1 : 0);
}
static void
ste_setmulti(sc)
struct ste_softc *sc;
{
struct ifnet *ifp;
int h = 0;
u_int32_t hashes[2] = { 0, 0 };
struct ifmultiaddr *ifma;
ifp = sc->ste_ifp;
if (ifp->if_flags & IFF_ALLMULTI || ifp->if_flags & IFF_PROMISC) {
STE_SETBIT1(sc, STE_RX_MODE, STE_RXMODE_ALLMULTI);
STE_CLRBIT1(sc, STE_RX_MODE, STE_RXMODE_MULTIHASH);
return;
}
/* first, zot all the existing hash bits */
CSR_WRITE_2(sc, STE_MAR0, 0);
CSR_WRITE_2(sc, STE_MAR1, 0);
CSR_WRITE_2(sc, STE_MAR2, 0);
CSR_WRITE_2(sc, STE_MAR3, 0);
/* now program new ones */
IF_ADDR_LOCK(ifp);
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
h = ether_crc32_be(LLADDR((struct sockaddr_dl *)
ifma->ifma_addr), ETHER_ADDR_LEN) & 0x3F;
if (h < 32)
hashes[0] |= (1 << h);
else
hashes[1] |= (1 << (h - 32));
}
IF_ADDR_UNLOCK(ifp);
CSR_WRITE_2(sc, STE_MAR0, hashes[0] & 0xFFFF);
CSR_WRITE_2(sc, STE_MAR1, (hashes[0] >> 16) & 0xFFFF);
CSR_WRITE_2(sc, STE_MAR2, hashes[1] & 0xFFFF);
CSR_WRITE_2(sc, STE_MAR3, (hashes[1] >> 16) & 0xFFFF);
STE_CLRBIT1(sc, STE_RX_MODE, STE_RXMODE_ALLMULTI);
STE_SETBIT1(sc, STE_RX_MODE, STE_RXMODE_MULTIHASH);
return;
}
#ifdef DEVICE_POLLING
static poll_handler_t ste_poll, ste_poll_locked;
static void
ste_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
struct ste_softc *sc = ifp->if_softc;
STE_LOCK(sc);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
ste_poll_locked(ifp, cmd, count);
STE_UNLOCK(sc);
}
static void
ste_poll_locked(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
struct ste_softc *sc = ifp->if_softc;
STE_LOCK_ASSERT(sc);
sc->rxcycles = count;
if (cmd == POLL_AND_CHECK_STATUS)
ste_rxeoc(sc);
ste_rxeof(sc);
ste_txeof(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ste_start_locked(ifp);
if (cmd == POLL_AND_CHECK_STATUS) {
u_int16_t status;
status = CSR_READ_2(sc, STE_ISR_ACK);
if (status & STE_ISR_TX_DONE)
ste_txeoc(sc);
if (status & STE_ISR_STATS_OFLOW) {
callout_stop(&sc->ste_stat_callout);
ste_stats_update(sc);
}
if (status & STE_ISR_LINKEVENT)
mii_pollstat(device_get_softc(sc->ste_miibus));
if (status & STE_ISR_HOSTERR) {
ste_reset(sc);
ste_init_locked(sc);
}
}
}
#endif /* DEVICE_POLLING */
static void
ste_intr(xsc)
void *xsc;
{
struct ste_softc *sc;
struct ifnet *ifp;
u_int16_t status;
sc = xsc;
STE_LOCK(sc);
ifp = sc->ste_ifp;
#ifdef DEVICE_POLLING
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
if (ifp->if_capenable & IFCAP_POLLING) {
STE_UNLOCK(sc);
return;
}
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
/* See if this is really our interrupt. */
if (!(CSR_READ_2(sc, STE_ISR) & STE_ISR_INTLATCH)) {
STE_UNLOCK(sc);
return;
}
for (;;) {
status = CSR_READ_2(sc, STE_ISR_ACK);
if (!(status & STE_INTRS))
break;
if (status & STE_ISR_RX_DMADONE) {
ste_rxeoc(sc);
ste_rxeof(sc);
}
if (status & STE_ISR_TX_DMADONE)
ste_txeof(sc);
if (status & STE_ISR_TX_DONE)
ste_txeoc(sc);
if (status & STE_ISR_STATS_OFLOW) {
callout_stop(&sc->ste_stat_callout);
ste_stats_update(sc);
}
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
if (status & STE_ISR_LINKEVENT)
mii_pollstat(device_get_softc(sc->ste_miibus));
if (status & STE_ISR_HOSTERR) {
ste_reset(sc);
ste_init_locked(sc);
}
}
/* Re-enable interrupts */
CSR_WRITE_2(sc, STE_IMR, STE_INTRS);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ste_start_locked(ifp);
STE_UNLOCK(sc);
return;
}
static void
ste_rxeoc(struct ste_softc *sc)
{
struct ste_chain_onefrag *cur_rx;
STE_LOCK_ASSERT(sc);
if (sc->ste_cdata.ste_rx_head->ste_ptr->ste_status == 0) {
cur_rx = sc->ste_cdata.ste_rx_head;
do {
cur_rx = cur_rx->ste_next;
/* If the ring is empty, just return. */
if (cur_rx == sc->ste_cdata.ste_rx_head)
return;
} while (cur_rx->ste_ptr->ste_status == 0);
if (sc->ste_cdata.ste_rx_head->ste_ptr->ste_status == 0) {
/* We've fallen behind the chip: catch it. */
sc->ste_cdata.ste_rx_head = cur_rx;
++ste_rxsyncs;
}
}
}
/*
* A frame has been uploaded: pass the resulting mbuf chain up to
* the higher level protocols.
*/
static void
ste_rxeof(sc)
struct ste_softc *sc;
{
struct mbuf *m;
struct ifnet *ifp;
struct ste_chain_onefrag *cur_rx;
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
int total_len = 0, count=0;
u_int32_t rxstat;
STE_LOCK_ASSERT(sc);
ifp = sc->ste_ifp;
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
while((rxstat = sc->ste_cdata.ste_rx_head->ste_ptr->ste_status)
& STE_RXSTAT_DMADONE) {
#ifdef DEVICE_POLLING
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
if (ifp->if_capenable & IFCAP_POLLING) {
if (sc->rxcycles <= 0)
break;
sc->rxcycles--;
}
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
if ((STE_RX_LIST_CNT - count) < 3) {
break;
}
cur_rx = sc->ste_cdata.ste_rx_head;
sc->ste_cdata.ste_rx_head = cur_rx->ste_next;
/*
* If an error occurs, update stats, clear the
* status word and leave the mbuf cluster in place:
* it should simply get re-used next time this descriptor
* comes up in the ring.
*/
if (rxstat & STE_RXSTAT_FRAME_ERR) {
ifp->if_ierrors++;
cur_rx->ste_ptr->ste_status = 0;
continue;
}
/*
* If there error bit was not set, the upload complete
* bit should be set which means we have a valid packet.
* If not, something truly strange has happened.
*/
if (!(rxstat & STE_RXSTAT_DMADONE)) {
device_printf(sc->ste_dev,
"bad receive status -- packet dropped\n");
ifp->if_ierrors++;
cur_rx->ste_ptr->ste_status = 0;
continue;
}
/* No errors; receive the packet. */
m = cur_rx->ste_mbuf;
total_len = cur_rx->ste_ptr->ste_status & STE_RXSTAT_FRAMELEN;
/*
* Try to conjure up a new mbuf cluster. If that
* fails, it means we have an out of memory condition and
* should leave the buffer in place and continue. This will
* result in a lost packet, but there's little else we
* can do in this situation.
*/
if (ste_newbuf(sc, cur_rx, NULL) == ENOBUFS) {
ifp->if_ierrors++;
cur_rx->ste_ptr->ste_status = 0;
continue;
}
m->m_pkthdr.rcvif = ifp;
m->m_pkthdr.len = m->m_len = total_len;
ifp->if_ipackets++;
STE_UNLOCK(sc);
(*ifp->if_input)(ifp, m);
STE_LOCK(sc);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
cur_rx->ste_ptr->ste_status = 0;
count++;
}
return;
}
static void
ste_txeoc(sc)
struct ste_softc *sc;
{
u_int8_t txstat;
struct ifnet *ifp;
ifp = sc->ste_ifp;
while ((txstat = CSR_READ_1(sc, STE_TX_STATUS)) &
STE_TXSTATUS_TXDONE) {
if (txstat & STE_TXSTATUS_UNDERRUN ||
txstat & STE_TXSTATUS_EXCESSCOLLS ||
txstat & STE_TXSTATUS_RECLAIMERR) {
ifp->if_oerrors++;
device_printf(sc->ste_dev,
"transmission error: %x\n", txstat);
ste_reset(sc);
ste_init_locked(sc);
if (txstat & STE_TXSTATUS_UNDERRUN &&
sc->ste_tx_thresh < STE_PACKET_SIZE) {
sc->ste_tx_thresh += STE_MIN_FRAMELEN;
device_printf(sc->ste_dev,
"tx underrun, increasing tx"
" start threshold to %d bytes\n",
sc->ste_tx_thresh);
}
CSR_WRITE_2(sc, STE_TX_STARTTHRESH, sc->ste_tx_thresh);
CSR_WRITE_2(sc, STE_TX_RECLAIM_THRESH,
(STE_PACKET_SIZE >> 4));
}
ste_init_locked(sc);
CSR_WRITE_2(sc, STE_TX_STATUS, txstat);
}
return;
}
static void
ste_txeof(sc)
struct ste_softc *sc;
{
struct ste_chain *cur_tx;
struct ifnet *ifp;
int idx;
ifp = sc->ste_ifp;
idx = sc->ste_cdata.ste_tx_cons;
while(idx != sc->ste_cdata.ste_tx_prod) {
cur_tx = &sc->ste_cdata.ste_tx_chain[idx];
if (!(cur_tx->ste_ptr->ste_ctl & STE_TXCTL_DMADONE))
break;
m_freem(cur_tx->ste_mbuf);
cur_tx->ste_mbuf = NULL;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
ifp->if_opackets++;
STE_INC(idx, STE_TX_LIST_CNT);
}
sc->ste_cdata.ste_tx_cons = idx;
if (idx == sc->ste_cdata.ste_tx_prod)
ifp->if_timer = 0;
}
static void
ste_stats_update(xsc)
void *xsc;
{
struct ste_softc *sc;
struct ifnet *ifp;
struct mii_data *mii;
sc = xsc;
STE_LOCK_ASSERT(sc);
ifp = sc->ste_ifp;
mii = device_get_softc(sc->ste_miibus);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
ifp->if_collisions += CSR_READ_1(sc, STE_LATE_COLLS)
+ CSR_READ_1(sc, STE_MULTI_COLLS)
+ CSR_READ_1(sc, STE_SINGLE_COLLS);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
if (!sc->ste_link) {
mii_pollstat(mii);
if (mii->mii_media_status & IFM_ACTIVE &&
IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
sc->ste_link++;
/*
* we don't get a call-back on re-init so do it
* otherwise we get stuck in the wrong link state
*/
ste_miibus_statchg(sc->ste_dev);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ste_start_locked(ifp);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
}
}
callout_reset(&sc->ste_stat_callout, hz, ste_stats_update, sc);
return;
}
/*
* Probe for a Sundance ST201 chip. Check the PCI vendor and device
* IDs against our list and return a device name if we find a match.
*/
static int
ste_probe(dev)
device_t dev;
{
struct ste_type *t;
t = ste_devs;
while(t->ste_name != NULL) {
if ((pci_get_vendor(dev) == t->ste_vid) &&
(pci_get_device(dev) == t->ste_did)) {
device_set_desc(dev, t->ste_name);
2005-02-24 21:32:56 +00:00
return (BUS_PROBE_DEFAULT);
}
t++;
}
return(ENXIO);
}
/*
* Attach the interface. Allocate softc structures, do ifmedia
* setup and ethernet/BPF attach.
*/
static int
ste_attach(dev)
device_t dev;
{
struct ste_softc *sc;
struct ifnet *ifp;
int error = 0, rid;
u_char eaddr[6];
sc = device_get_softc(dev);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
sc->ste_dev = dev;
/*
* Only use one PHY since this chip reports multiple
* Note on the DFE-550 the PHY is at 1 on the DFE-580
* it is at 0 & 1. It is rev 0x12.
*/
if (pci_get_vendor(dev) == DL_VENDORID &&
pci_get_device(dev) == DL_DEVICEID_DL10050 &&
pci_get_revid(dev) == 0x12 )
sc->ste_one_phy = 1;
mtx_init(&sc->ste_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
MTX_DEF);
/*
* Map control/status registers.
*/
pci_enable_busmaster(dev);
rid = STE_RID;
sc->ste_res = bus_alloc_resource_any(dev, STE_RES, &rid, RF_ACTIVE);
if (sc->ste_res == NULL) {
device_printf(dev, "couldn't map ports/memory\n");
error = ENXIO;
goto fail;
}
sc->ste_btag = rman_get_bustag(sc->ste_res);
sc->ste_bhandle = rman_get_bushandle(sc->ste_res);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
/* Allocate interrupt */
rid = 0;
sc->ste_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
RF_SHAREABLE | RF_ACTIVE);
if (sc->ste_irq == NULL) {
device_printf(dev, "couldn't map interrupt\n");
error = ENXIO;
goto fail;
}
callout_init_mtx(&sc->ste_stat_callout, &sc->ste_mtx, 0);
/* Reset the adapter. */
ste_reset(sc);
/*
* Get station address from the EEPROM.
*/
if (ste_read_eeprom(sc, eaddr,
STE_EEADDR_NODE0, 3, 0)) {
device_printf(dev, "failed to read station address\n");
error = ENXIO;;
goto fail;
}
/* Allocate the descriptor queues. */
sc->ste_ldata = contigmalloc(sizeof(struct ste_list_data), M_DEVBUF,
M_NOWAIT, 0, 0xffffffff, PAGE_SIZE, 0);
if (sc->ste_ldata == NULL) {
device_printf(dev, "no memory for list buffers!\n");
error = ENXIO;
goto fail;
}
bzero(sc->ste_ldata, sizeof(struct ste_list_data));
ifp = sc->ste_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
device_printf(dev, "can not if_alloc()\n");
error = ENOSPC;
goto fail;
}
/* Do MII setup. */
if (mii_phy_probe(dev, &sc->ste_miibus,
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
ste_ifmedia_upd, ste_ifmedia_sts)) {
device_printf(dev, "MII without any phy!\n");
error = ENXIO;
goto fail;
}
ifp->if_softc = sc;
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
ifp->if_mtu = ETHERMTU;
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = ste_ioctl;
ifp->if_start = ste_start;
ifp->if_watchdog = ste_watchdog;
ifp->if_init = ste_init;
IFQ_SET_MAXLEN(&ifp->if_snd, STE_TX_LIST_CNT - 1);
ifp->if_snd.ifq_drv_maxlen = STE_TX_LIST_CNT - 1;
IFQ_SET_READY(&ifp->if_snd);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
sc->ste_tx_thresh = STE_TXSTART_THRESH;
/*
* Call MI attach routine.
*/
ether_ifattach(ifp, eaddr);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/*
* Tell the upper layer(s) we support long frames.
*/
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
ifp->if_capabilities |= IFCAP_VLAN_MTU;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
ifp->if_capenable = ifp->if_capabilities;
#ifdef DEVICE_POLLING
ifp->if_capabilities |= IFCAP_POLLING;
#endif
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/* Hook interrupt last to avoid having to lock softc */
error = bus_setup_intr(dev, sc->ste_irq, INTR_TYPE_NET | INTR_MPSAFE,
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
ste_intr, sc, &sc->ste_intrhand);
if (error) {
device_printf(dev, "couldn't set up irq\n");
ether_ifdetach(ifp);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
goto fail;
}
fail:
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (error)
ste_detach(dev);
return(error);
}
/*
* Shutdown hardware and free up resources. This can be called any
* time after the mutex has been initialized. It is called in both
* the error case in attach and the normal detach case so it needs
* to be careful about only freeing resources that have actually been
* allocated.
*/
static int
ste_detach(dev)
device_t dev;
{
struct ste_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
KASSERT(mtx_initialized(&sc->ste_mtx), ("ste mutex not initialized"));
ifp = sc->ste_ifp;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING)
ether_poll_deregister(ifp);
#endif
/* These should only be active if attach succeeded */
if (device_is_attached(dev)) {
STE_LOCK(sc);
ste_stop(sc);
STE_UNLOCK(sc);
callout_drain(&sc->ste_stat_callout);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
ether_ifdetach(ifp);
}
if (sc->ste_miibus)
device_delete_child(dev, sc->ste_miibus);
bus_generic_detach(dev);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (sc->ste_intrhand)
bus_teardown_intr(dev, sc->ste_irq, sc->ste_intrhand);
if (sc->ste_irq)
bus_release_resource(dev, SYS_RES_IRQ, 0, sc->ste_irq);
if (sc->ste_res)
bus_release_resource(dev, STE_RES, STE_RID, sc->ste_res);
if (ifp)
if_free(ifp);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (sc->ste_ldata) {
contigfree(sc->ste_ldata, sizeof(struct ste_list_data),
M_DEVBUF);
}
mtx_destroy(&sc->ste_mtx);
return(0);
}
static int
ste_newbuf(sc, c, m)
struct ste_softc *sc;
struct ste_chain_onefrag *c;
struct mbuf *m;
{
struct mbuf *m_new = NULL;
if (m == NULL) {
MGETHDR(m_new, M_DONTWAIT, MT_DATA);
if (m_new == NULL)
return(ENOBUFS);
MCLGET(m_new, M_DONTWAIT);
if (!(m_new->m_flags & M_EXT)) {
m_freem(m_new);
return(ENOBUFS);
}
m_new->m_len = m_new->m_pkthdr.len = MCLBYTES;
} else {
m_new = m;
m_new->m_len = m_new->m_pkthdr.len = MCLBYTES;
m_new->m_data = m_new->m_ext.ext_buf;
}
m_adj(m_new, ETHER_ALIGN);
c->ste_mbuf = m_new;
c->ste_ptr->ste_status = 0;
c->ste_ptr->ste_frag.ste_addr = vtophys(mtod(m_new, caddr_t));
c->ste_ptr->ste_frag.ste_len = (1536 + ETHER_VLAN_ENCAP_LEN) | STE_FRAG_LAST;
return(0);
}
static int
ste_init_rx_list(sc)
struct ste_softc *sc;
{
struct ste_chain_data *cd;
struct ste_list_data *ld;
int i;
cd = &sc->ste_cdata;
ld = sc->ste_ldata;
for (i = 0; i < STE_RX_LIST_CNT; i++) {
cd->ste_rx_chain[i].ste_ptr = &ld->ste_rx_list[i];
if (ste_newbuf(sc, &cd->ste_rx_chain[i], NULL) == ENOBUFS)
return(ENOBUFS);
if (i == (STE_RX_LIST_CNT - 1)) {
cd->ste_rx_chain[i].ste_next =
&cd->ste_rx_chain[0];
ld->ste_rx_list[i].ste_next =
vtophys(&ld->ste_rx_list[0]);
} else {
cd->ste_rx_chain[i].ste_next =
&cd->ste_rx_chain[i + 1];
ld->ste_rx_list[i].ste_next =
vtophys(&ld->ste_rx_list[i + 1]);
}
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
ld->ste_rx_list[i].ste_status = 0;
}
cd->ste_rx_head = &cd->ste_rx_chain[0];
return(0);
}
static void
ste_init_tx_list(sc)
struct ste_softc *sc;
{
struct ste_chain_data *cd;
struct ste_list_data *ld;
int i;
cd = &sc->ste_cdata;
ld = sc->ste_ldata;
for (i = 0; i < STE_TX_LIST_CNT; i++) {
cd->ste_tx_chain[i].ste_ptr = &ld->ste_tx_list[i];
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
cd->ste_tx_chain[i].ste_ptr->ste_next = 0;
cd->ste_tx_chain[i].ste_ptr->ste_ctl = 0;
cd->ste_tx_chain[i].ste_phys = vtophys(&ld->ste_tx_list[i]);
if (i == (STE_TX_LIST_CNT - 1))
cd->ste_tx_chain[i].ste_next =
&cd->ste_tx_chain[0];
else
cd->ste_tx_chain[i].ste_next =
&cd->ste_tx_chain[i + 1];
}
cd->ste_tx_prod = 0;
cd->ste_tx_cons = 0;
return;
}
static void
ste_init(xsc)
void *xsc;
{
struct ste_softc *sc;
sc = xsc;
STE_LOCK(sc);
ste_init_locked(sc);
STE_UNLOCK(sc);
}
static void
ste_init_locked(sc)
struct ste_softc *sc;
{
int i;
struct ifnet *ifp;
STE_LOCK_ASSERT(sc);
ifp = sc->ste_ifp;
ste_stop(sc);
/* Init our MAC address */
for (i = 0; i < ETHER_ADDR_LEN; i++) {
CSR_WRITE_1(sc, STE_PAR0 + i, IF_LLADDR(sc->ste_ifp)[i]);
}
/* Init RX list */
if (ste_init_rx_list(sc) == ENOBUFS) {
device_printf(sc->ste_dev,
"initialization failed: no memory for RX buffers\n");
ste_stop(sc);
return;
}
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/* Set RX polling interval */
CSR_WRITE_1(sc, STE_RX_DMAPOLL_PERIOD, 64);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/* Init TX descriptors */
ste_init_tx_list(sc);
/* Set the TX freethresh value */
CSR_WRITE_1(sc, STE_TX_DMABURST_THRESH, STE_PACKET_SIZE >> 8);
/* Set the TX start threshold for best performance. */
CSR_WRITE_2(sc, STE_TX_STARTTHRESH, sc->ste_tx_thresh);
/* Set the TX reclaim threshold. */
CSR_WRITE_1(sc, STE_TX_RECLAIM_THRESH, (STE_PACKET_SIZE >> 4));
/* Set up the RX filter. */
CSR_WRITE_1(sc, STE_RX_MODE, STE_RXMODE_UNICAST);
/* If we want promiscuous mode, set the allframes bit. */
if (ifp->if_flags & IFF_PROMISC) {
STE_SETBIT1(sc, STE_RX_MODE, STE_RXMODE_PROMISC);
} else {
STE_CLRBIT1(sc, STE_RX_MODE, STE_RXMODE_PROMISC);
}
/* Set capture broadcast bit to accept broadcast frames. */
if (ifp->if_flags & IFF_BROADCAST) {
STE_SETBIT1(sc, STE_RX_MODE, STE_RXMODE_BROADCAST);
} else {
STE_CLRBIT1(sc, STE_RX_MODE, STE_RXMODE_BROADCAST);
}
ste_setmulti(sc);
/* Load the address of the RX list. */
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_RXDMA_STALL);
ste_wait(sc);
CSR_WRITE_4(sc, STE_RX_DMALIST_PTR,
vtophys(&sc->ste_ldata->ste_rx_list[0]));
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_RXDMA_UNSTALL);
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_RXDMA_UNSTALL);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/* Set TX polling interval (defer until we TX first packet */
CSR_WRITE_1(sc, STE_TX_DMAPOLL_PERIOD, 0);
/* Load address of the TX list */
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_TXDMA_STALL);
ste_wait(sc);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
CSR_WRITE_4(sc, STE_TX_DMALIST_PTR, 0);
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_TXDMA_UNSTALL);
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_TXDMA_UNSTALL);
ste_wait(sc);
sc->ste_tx_prev = NULL;
/* Enable receiver and transmitter */
CSR_WRITE_2(sc, STE_MACCTL0, 0);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
CSR_WRITE_2(sc, STE_MACCTL1, 0);
STE_SETBIT2(sc, STE_MACCTL1, STE_MACCTL1_TX_ENABLE);
STE_SETBIT2(sc, STE_MACCTL1, STE_MACCTL1_RX_ENABLE);
/* Enable stats counters. */
STE_SETBIT2(sc, STE_MACCTL1, STE_MACCTL1_STATS_ENABLE);
CSR_WRITE_2(sc, STE_ISR, 0xFFFF);
#ifdef DEVICE_POLLING
/* Disable interrupts if we are polling. */
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
if (ifp->if_capenable & IFCAP_POLLING)
CSR_WRITE_2(sc, STE_IMR, 0);
else
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
/* Enable interrupts. */
CSR_WRITE_2(sc, STE_IMR, STE_INTRS);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/* Accept VLAN length packets */
CSR_WRITE_2(sc, STE_MAX_FRAMELEN, ETHER_MAX_LEN + ETHER_VLAN_ENCAP_LEN);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
ste_ifmedia_upd_locked(ifp);
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
callout_reset(&sc->ste_stat_callout, hz, ste_stats_update, sc);
return;
}
static void
ste_stop(sc)
struct ste_softc *sc;
{
int i;
struct ifnet *ifp;
STE_LOCK_ASSERT(sc);
ifp = sc->ste_ifp;
callout_stop(&sc->ste_stat_callout);
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING|IFF_DRV_OACTIVE);
CSR_WRITE_2(sc, STE_IMR, 0);
STE_SETBIT2(sc, STE_MACCTL1, STE_MACCTL1_TX_DISABLE);
STE_SETBIT2(sc, STE_MACCTL1, STE_MACCTL1_RX_DISABLE);
STE_SETBIT2(sc, STE_MACCTL1, STE_MACCTL1_STATS_DISABLE);
STE_SETBIT2(sc, STE_DMACTL, STE_DMACTL_TXDMA_STALL);
STE_SETBIT2(sc, STE_DMACTL, STE_DMACTL_RXDMA_STALL);
ste_wait(sc);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
/*
* Try really hard to stop the RX engine or under heavy RX
* data chip will write into de-allocated memory.
*/
ste_reset(sc);
sc->ste_link = 0;
for (i = 0; i < STE_RX_LIST_CNT; i++) {
if (sc->ste_cdata.ste_rx_chain[i].ste_mbuf != NULL) {
m_freem(sc->ste_cdata.ste_rx_chain[i].ste_mbuf);
sc->ste_cdata.ste_rx_chain[i].ste_mbuf = NULL;
}
}
for (i = 0; i < STE_TX_LIST_CNT; i++) {
if (sc->ste_cdata.ste_tx_chain[i].ste_mbuf != NULL) {
m_freem(sc->ste_cdata.ste_tx_chain[i].ste_mbuf);
sc->ste_cdata.ste_tx_chain[i].ste_mbuf = NULL;
}
}
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
bzero(sc->ste_ldata, sizeof(struct ste_list_data));
return;
}
static void
ste_reset(sc)
struct ste_softc *sc;
{
int i;
STE_SETBIT4(sc, STE_ASICCTL,
STE_ASICCTL_GLOBAL_RESET|STE_ASICCTL_RX_RESET|
STE_ASICCTL_TX_RESET|STE_ASICCTL_DMA_RESET|
STE_ASICCTL_FIFO_RESET|STE_ASICCTL_NETWORK_RESET|
STE_ASICCTL_AUTOINIT_RESET|STE_ASICCTL_HOST_RESET|
STE_ASICCTL_EXTRESET_RESET);
DELAY(100000);
for (i = 0; i < STE_TIMEOUT; i++) {
if (!(CSR_READ_4(sc, STE_ASICCTL) & STE_ASICCTL_RESET_BUSY))
break;
}
if (i == STE_TIMEOUT)
device_printf(sc->ste_dev, "global reset never completed\n");
return;
}
static int
ste_ioctl(ifp, command, data)
struct ifnet *ifp;
u_long command;
caddr_t data;
{
struct ste_softc *sc;
struct ifreq *ifr;
struct mii_data *mii;
int error = 0;
sc = ifp->if_softc;
ifr = (struct ifreq *)data;
switch(command) {
case SIOCSIFFLAGS:
STE_LOCK(sc);
if (ifp->if_flags & IFF_UP) {
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
ifp->if_flags & IFF_PROMISC &&
!(sc->ste_if_flags & IFF_PROMISC)) {
STE_SETBIT1(sc, STE_RX_MODE,
STE_RXMODE_PROMISC);
} else if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
!(ifp->if_flags & IFF_PROMISC) &&
sc->ste_if_flags & IFF_PROMISC) {
STE_CLRBIT1(sc, STE_RX_MODE,
STE_RXMODE_PROMISC);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
}
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
(ifp->if_flags ^ sc->ste_if_flags) & IFF_ALLMULTI)
ste_setmulti(sc);
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
sc->ste_tx_thresh = STE_TXSTART_THRESH;
ste_init_locked(sc);
}
} else {
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
ste_stop(sc);
}
sc->ste_if_flags = ifp->if_flags;
STE_UNLOCK(sc);
error = 0;
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
STE_LOCK(sc);
ste_setmulti(sc);
STE_UNLOCK(sc);
error = 0;
break;
case SIOCGIFMEDIA:
case SIOCSIFMEDIA:
mii = device_get_softc(sc->ste_miibus);
error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
break;
case SIOCSIFCAP:
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#ifdef DEVICE_POLLING
if (ifr->ifr_reqcap & IFCAP_POLLING &&
!(ifp->if_capenable & IFCAP_POLLING)) {
error = ether_poll_register(ste_poll, ifp);
if (error)
return(error);
STE_LOCK(sc);
/* Disable interrupts */
CSR_WRITE_2(sc, STE_IMR, 0);
ifp->if_capenable |= IFCAP_POLLING;
STE_UNLOCK(sc);
return (error);
}
if (!(ifr->ifr_reqcap & IFCAP_POLLING) &&
ifp->if_capenable & IFCAP_POLLING) {
error = ether_poll_deregister(ifp);
/* Enable interrupts. */
STE_LOCK(sc);
CSR_WRITE_2(sc, STE_IMR, STE_INTRS);
ifp->if_capenable &= ~IFCAP_POLLING;
STE_UNLOCK(sc);
return (error);
}
#endif /* DEVICE_POLLING */
break;
default:
error = ether_ioctl(ifp, command, data);
break;
}
return(error);
}
static int
ste_encap(sc, c, m_head)
struct ste_softc *sc;
struct ste_chain *c;
struct mbuf *m_head;
{
int frag = 0;
struct ste_frag *f = NULL;
struct mbuf *m;
struct ste_desc *d;
d = c->ste_ptr;
d->ste_ctl = 0;
encap_retry:
for (m = m_head, frag = 0; m != NULL; m = m->m_next) {
if (m->m_len != 0) {
if (frag == STE_MAXFRAGS)
break;
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
f = &d->ste_frags[frag];
f->ste_addr = vtophys(mtod(m, vm_offset_t));
f->ste_len = m->m_len;
frag++;
}
}
if (m != NULL) {
struct mbuf *mn;
/*
* We ran out of segments. We have to recopy this
* mbuf chain first. Bail out if we can't get the
* new buffers.
*/
mn = m_defrag(m_head, M_DONTWAIT);
if (mn == NULL) {
m_freem(m_head);
return ENOMEM;
}
m_head = mn;
goto encap_retry;
}
c->ste_mbuf = m_head;
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
d->ste_frags[frag - 1].ste_len |= STE_FRAG_LAST;
d->ste_ctl = 1;
return(0);
}
static void
ste_start(ifp)
struct ifnet *ifp;
{
struct ste_softc *sc;
sc = ifp->if_softc;
STE_LOCK(sc);
ste_start_locked(ifp);
STE_UNLOCK(sc);
}
static void
ste_start_locked(ifp)
struct ifnet *ifp;
{
struct ste_softc *sc;
struct mbuf *m_head = NULL;
struct ste_chain *cur_tx;
int idx;
sc = ifp->if_softc;
STE_LOCK_ASSERT(sc);
if (!sc->ste_link)
return;
if (ifp->if_drv_flags & IFF_DRV_OACTIVE)
return;
idx = sc->ste_cdata.ste_tx_prod;
while(sc->ste_cdata.ste_tx_chain[idx].ste_mbuf == NULL) {
/*
* We cannot re-use the last (free) descriptor;
* the chip may not have read its ste_next yet.
*/
if (STE_NEXT(idx, STE_TX_LIST_CNT) ==
sc->ste_cdata.ste_tx_cons) {
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
break;
}
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
if (m_head == NULL)
break;
cur_tx = &sc->ste_cdata.ste_tx_chain[idx];
if (ste_encap(sc, cur_tx, m_head) != 0)
break;
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
cur_tx->ste_ptr->ste_next = 0;
if (sc->ste_tx_prev == NULL) {
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
cur_tx->ste_ptr->ste_ctl = STE_TXCTL_DMAINTR | 1;
/* Load address of the TX list */
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_TXDMA_STALL);
ste_wait(sc);
CSR_WRITE_4(sc, STE_TX_DMALIST_PTR,
vtophys(&sc->ste_ldata->ste_tx_list[0]));
/* Set TX polling interval to start TX engine */
CSR_WRITE_1(sc, STE_TX_DMAPOLL_PERIOD, 64);
STE_SETBIT4(sc, STE_DMACTL, STE_DMACTL_TXDMA_UNSTALL);
ste_wait(sc);
}else{
cur_tx->ste_ptr->ste_ctl = STE_TXCTL_DMAINTR | 1;
sc->ste_tx_prev->ste_ptr->ste_next
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
= cur_tx->ste_phys;
}
sc->ste_tx_prev = cur_tx;
/*
* If there's a BPF listener, bounce a copy of this frame
* to him.
*/
BPF_MTAP(ifp, cur_tx->ste_mbuf);
STE_INC(idx, STE_TX_LIST_CNT);
Fixes for the D-Link DFE-580 card. This is pretty much fixes any issue I can find: - Watchdog timeouts were due to starting the TX DMA engine before we had a packet ready for it. So the first packet sent never got out only if we sent more then one packet at a time did the others make it out and not blow up. Of course reseting the chip then caused us not to transmit the first packet again ie. catch-22. This required logic changes. - Combine interrupts on TX packets being queued up. - Don't keep running around the RX ring since we might get out of sync so only go around once per receive - Let the RX engine recover via the poll interface which is similar to the TX interface. This way the chip wakes up with no effort when we read enough packets. - Do better hand-shaking on RX & TX packets so they don't start of to soon. - Force a duplex setting when the link comes up after an ste_init or it will default to half-duplex and be really slow. This only happens on subsequent ste_init. The first one worked. - Don't call stat_update for every overflow. We only monitor the collisions so the tick interval is good enough for that. Just read in the collision stats to minimize bus reads. - Don't read the miibus every tick since it uses delays and delays are not good for performance. - Tie link events directly to the miibus code so the port gets set correctly if someone changes the port settings. - Reduce the extreme number of {R,T}FD's. They would consume 130K of kernel memory for each NIC. - Set the TX_THRESH to wait for the DMA engine to complete before running the TX FIFO. This hurts peak TX performance but under bi-directional load the DMA engine can't keep up with the FIFO. Testing shows that we end up in the case anyways (a la dc(4) issues but worse since the RX engine hogs everything). - When stopping the card do a reset since the reset verifies the card has stopped. Otherwise on heavy RX load the RX DMA engine is still stuffing packets into memory. If that happens after we free the DMA area memory bits get scribled in memory and bad things happen. This card still has seemingly unfixable issues under heavy RX load in which the card takes over the PCI bus. Sponsored by: Vernier Networks MFC after: 1 week
2002-08-07 22:31:27 +00:00
ifp->if_timer = 5;
}
sc->ste_cdata.ste_tx_prod = idx;
return;
}
static void
ste_watchdog(ifp)
struct ifnet *ifp;
{
struct ste_softc *sc;
sc = ifp->if_softc;
STE_LOCK(sc);
ifp->if_oerrors++;
if_printf(ifp, "watchdog timeout\n");
ste_txeoc(sc);
ste_txeof(sc);
ste_rxeoc(sc);
ste_rxeof(sc);
ste_reset(sc);
ste_init_locked(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ste_start_locked(ifp);
STE_UNLOCK(sc);
return;
}
static void
ste_shutdown(dev)
device_t dev;
{
struct ste_softc *sc;
sc = device_get_softc(dev);
STE_LOCK(sc);
ste_stop(sc);
STE_UNLOCK(sc);
return;
}