linux/drivers/net/sun3lance.c

961 lines
26 KiB
C
Raw Normal View History

/* sun3lance.c: Ethernet driver for SUN3 Lance chip */
/*
Sun3 Lance ethernet driver, by Sam Creasey (sammy@users.qual.net).
This driver is a part of the linux kernel, and is thus distributed
under the GNU General Public License.
The values used in LANCE_OBIO and LANCE_IRQ seem to be empirically
true for the correct IRQ and address of the lance registers. They
have not been widely tested, however. What we probably need is a
"proper" way to search for a device in the sun3's prom, but, alas,
linux has no such thing.
This driver is largely based on atarilance.c, by Roman Hodek. Other
sources of inspiration were the NetBSD sun3 am7990 driver, and the
linux sparc lance driver (sunlance.c).
There are more assumptions made throughout this driver, it almost
certainly still needs work, but it does work at least for RARP/BOOTP and
mounting the root NFS filesystem.
*/
static char *version = "sun3lance.c: v1.2 1/12/2001 Sam Creasey (sammy@sammy.net)\n";
#include <linux/module.h>
#include <linux/stddef.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/interrupt.h>
#include <linux/init.h>
#include <linux/ioport.h>
#include <linux/delay.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/bitops.h>
#include <asm/cacheflush.h>
#include <asm/setup.h>
#include <asm/irq.h>
#include <asm/io.h>
#include <asm/pgtable.h>
#include <asm/dvma.h>
#include <asm/idprom.h>
#include <asm/machines.h>
#ifdef CONFIG_SUN3
#include <asm/sun3mmu.h>
#else
#include <asm/sun3xprom.h>
#endif
/* sun3/60 addr/irq for the lance chip. If your sun is different,
change this. */
#define LANCE_OBIO 0x120000
#define LANCE_IRQ IRQ_AUTO_3
/* Debug level:
* 0 = silent, print only serious errors
* 1 = normal, print error messages
* 2 = debug, print debug infos
* 3 = debug, print even more debug infos (packet data)
*/
#define LANCE_DEBUG 0
#ifdef LANCE_DEBUG
static int lance_debug = LANCE_DEBUG;
#else
static int lance_debug = 1;
#endif
module_param(lance_debug, int, 0);
MODULE_PARM_DESC(lance_debug, "SUN3 Lance debug level (0-3)");
MODULE_LICENSE("GPL");
#define DPRINTK(n,a) \
do { \
if (lance_debug >= n) \
printk a; \
} while( 0 )
/* we're only using 32k of memory, so we use 4 TX
buffers and 16 RX buffers. These values are expressed as log2. */
#define TX_LOG_RING_SIZE 3
#define RX_LOG_RING_SIZE 5
/* These are the derived values */
#define TX_RING_SIZE (1 << TX_LOG_RING_SIZE)
#define TX_RING_LEN_BITS (TX_LOG_RING_SIZE << 5)
#define TX_RING_MOD_MASK (TX_RING_SIZE - 1)
#define RX_RING_SIZE (1 << RX_LOG_RING_SIZE)
#define RX_RING_LEN_BITS (RX_LOG_RING_SIZE << 5)
#define RX_RING_MOD_MASK (RX_RING_SIZE - 1)
/* Definitions for packet buffer access: */
#define PKT_BUF_SZ 1544
/* Get the address of a packet buffer corresponding to a given buffer head */
#define PKTBUF_ADDR(head) (void *)((unsigned long)(MEM) | (head)->base)
/* The LANCE Rx and Tx ring descriptors. */
struct lance_rx_head {
unsigned short base; /* Low word of base addr */
volatile unsigned char flag;
unsigned char base_hi; /* High word of base addr (unused) */
short buf_length; /* This length is 2s complement! */
volatile short msg_length; /* This length is "normal". */
};
struct lance_tx_head {
unsigned short base; /* Low word of base addr */
volatile unsigned char flag;
unsigned char base_hi; /* High word of base addr (unused) */
short length; /* Length is 2s complement! */
volatile short misc;
};
/* The LANCE initialization block, described in databook. */
struct lance_init_block {
unsigned short mode; /* Pre-set mode */
unsigned char hwaddr[6]; /* Physical ethernet address */
unsigned int filter[2]; /* Multicast filter (unused). */
/* Receive and transmit ring base, along with length bits. */
unsigned short rdra;
unsigned short rlen;
unsigned short tdra;
unsigned short tlen;
unsigned short pad[4]; /* is thie needed? */
};
/* The whole layout of the Lance shared memory */
struct lance_memory {
struct lance_init_block init;
struct lance_tx_head tx_head[TX_RING_SIZE];
struct lance_rx_head rx_head[RX_RING_SIZE];
char rx_data[RX_RING_SIZE][PKT_BUF_SZ];
char tx_data[TX_RING_SIZE][PKT_BUF_SZ];
};
/* The driver's private device structure */
struct lance_private {
volatile unsigned short *iobase;
struct lance_memory *mem;
int new_rx, new_tx; /* The next free ring entry */
int old_tx, old_rx; /* ring entry to be processed */
/* These two must be longs for set_bit() */
long tx_full;
long lock;
};
/* I/O register access macros */
#define MEM lp->mem
#define DREG lp->iobase[0]
#define AREG lp->iobase[1]
#define REGA(a) (*( AREG = (a), &DREG ))
/* Definitions for the Lance */
/* tx_head flags */
#define TMD1_ENP 0x01 /* end of packet */
#define TMD1_STP 0x02 /* start of packet */
#define TMD1_DEF 0x04 /* deferred */
#define TMD1_ONE 0x08 /* one retry needed */
#define TMD1_MORE 0x10 /* more than one retry needed */
#define TMD1_ERR 0x40 /* error summary */
#define TMD1_OWN 0x80 /* ownership (set: chip owns) */
#define TMD1_OWN_CHIP TMD1_OWN
#define TMD1_OWN_HOST 0
/* tx_head misc field */
#define TMD3_TDR 0x03FF /* Time Domain Reflectometry counter */
#define TMD3_RTRY 0x0400 /* failed after 16 retries */
#define TMD3_LCAR 0x0800 /* carrier lost */
#define TMD3_LCOL 0x1000 /* late collision */
#define TMD3_UFLO 0x4000 /* underflow (late memory) */
#define TMD3_BUFF 0x8000 /* buffering error (no ENP) */
/* rx_head flags */
#define RMD1_ENP 0x01 /* end of packet */
#define RMD1_STP 0x02 /* start of packet */
#define RMD1_BUFF 0x04 /* buffer error */
#define RMD1_CRC 0x08 /* CRC error */
#define RMD1_OFLO 0x10 /* overflow */
#define RMD1_FRAM 0x20 /* framing error */
#define RMD1_ERR 0x40 /* error summary */
#define RMD1_OWN 0x80 /* ownership (set: ship owns) */
#define RMD1_OWN_CHIP RMD1_OWN
#define RMD1_OWN_HOST 0
/* register names */
#define CSR0 0 /* mode/status */
#define CSR1 1 /* init block addr (low) */
#define CSR2 2 /* init block addr (high) */
#define CSR3 3 /* misc */
#define CSR8 8 /* address filter */
#define CSR15 15 /* promiscuous mode */
/* CSR0 */
/* (R=readable, W=writeable, S=set on write, C=clear on write) */
#define CSR0_INIT 0x0001 /* initialize (RS) */
#define CSR0_STRT 0x0002 /* start (RS) */
#define CSR0_STOP 0x0004 /* stop (RS) */
#define CSR0_TDMD 0x0008 /* transmit demand (RS) */
#define CSR0_TXON 0x0010 /* transmitter on (R) */
#define CSR0_RXON 0x0020 /* receiver on (R) */
#define CSR0_INEA 0x0040 /* interrupt enable (RW) */
#define CSR0_INTR 0x0080 /* interrupt active (R) */
#define CSR0_IDON 0x0100 /* initialization done (RC) */
#define CSR0_TINT 0x0200 /* transmitter interrupt (RC) */
#define CSR0_RINT 0x0400 /* receiver interrupt (RC) */
#define CSR0_MERR 0x0800 /* memory error (RC) */
#define CSR0_MISS 0x1000 /* missed frame (RC) */
#define CSR0_CERR 0x2000 /* carrier error (no heartbeat :-) (RC) */
#define CSR0_BABL 0x4000 /* babble: tx-ed too many bits (RC) */
#define CSR0_ERR 0x8000 /* error (RC) */
/* CSR3 */
#define CSR3_BCON 0x0001 /* byte control */
#define CSR3_ACON 0x0002 /* ALE control */
#define CSR3_BSWP 0x0004 /* byte swap (1=big endian) */
/***************************** Prototypes *****************************/
static int lance_probe( struct net_device *dev);
static int lance_open( struct net_device *dev );
static void lance_init_ring( struct net_device *dev );
static int lance_start_xmit( struct sk_buff *skb, struct net_device *dev );
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t lance_interrupt( int irq, void *dev_id);
static int lance_rx( struct net_device *dev );
static int lance_close( struct net_device *dev );
static void set_multicast_list( struct net_device *dev );
/************************* End of Prototypes **************************/
struct net_device * __init sun3lance_probe(int unit)
{
struct net_device *dev;
static int found;
int err = -ENODEV;
/* check that this machine has an onboard lance */
switch(idprom->id_machtype) {
case SM_SUN3|SM_3_50:
case SM_SUN3|SM_3_60:
case SM_SUN3X|SM_3_80:
/* these machines have lance */
break;
default:
return ERR_PTR(-ENODEV);
}
if (found)
return ERR_PTR(-ENODEV);
dev = alloc_etherdev(sizeof(struct lance_private));
if (!dev)
return ERR_PTR(-ENOMEM);
if (unit >= 0) {
sprintf(dev->name, "eth%d", unit);
netdev_boot_setup_check(dev);
}
if (!lance_probe(dev))
goto out;
err = register_netdev(dev);
if (err)
goto out1;
found = 1;
return dev;
out1:
#ifdef CONFIG_SUN3
iounmap((void __iomem *)dev->base_addr);
#endif
out:
free_netdev(dev);
return ERR_PTR(err);
}
static int __init lance_probe( struct net_device *dev)
{
unsigned long ioaddr;
struct lance_private *lp;
int i;
static int did_version;
volatile unsigned short *ioaddr_probe;
unsigned short tmp1, tmp2;
DECLARE_MAC_BUF(mac);
#ifdef CONFIG_SUN3
ioaddr = (unsigned long)ioremap(LANCE_OBIO, PAGE_SIZE);
if (!ioaddr)
return 0;
#else
ioaddr = SUN3X_LANCE;
#endif
/* test to see if there's really a lance here */
/* (CSRO_INIT shouldn't be readable) */
ioaddr_probe = (volatile unsigned short *)ioaddr;
tmp1 = ioaddr_probe[0];
tmp2 = ioaddr_probe[1];
ioaddr_probe[1] = CSR0;
ioaddr_probe[0] = CSR0_INIT | CSR0_STOP;
if(ioaddr_probe[0] != CSR0_STOP) {
ioaddr_probe[0] = tmp1;
ioaddr_probe[1] = tmp2;
#ifdef CONFIG_SUN3
iounmap((void __iomem *)ioaddr);
#endif
return 0;
}
lp = netdev_priv(dev);
/* XXX - leak? */
MEM = dvma_malloc_align(sizeof(struct lance_memory), 0x10000);
if (MEM == NULL) {
#ifdef CONFIG_SUN3
iounmap((void __iomem *)ioaddr);
#endif
printk(KERN_WARNING "SUN3 Lance couldn't allocate DVMA memory\n");
return 0;
}
lp->iobase = (volatile unsigned short *)ioaddr;
dev->base_addr = (unsigned long)ioaddr; /* informational only */
REGA(CSR0) = CSR0_STOP;
if (request_irq(LANCE_IRQ, lance_interrupt, IRQF_DISABLED, "SUN3 Lance", dev) < 0) {
#ifdef CONFIG_SUN3
iounmap((void __iomem *)ioaddr);
#endif
dvma_free((void *)MEM);
printk(KERN_WARNING "SUN3 Lance unable to allocate IRQ\n");
return 0;
}
dev->irq = (unsigned short)LANCE_IRQ;
printk("%s: SUN3 Lance at io %#lx, mem %#lx, irq %d, hwaddr ",
dev->name,
(unsigned long)ioaddr,
(unsigned long)MEM,
dev->irq);
/* copy in the ethernet address from the prom */
for(i = 0; i < 6 ; i++)
dev->dev_addr[i] = idprom->id_ethaddr[i];
/* tell the card it's ether address, bytes swapped */
MEM->init.hwaddr[0] = dev->dev_addr[1];
MEM->init.hwaddr[1] = dev->dev_addr[0];
MEM->init.hwaddr[2] = dev->dev_addr[3];
MEM->init.hwaddr[3] = dev->dev_addr[2];
MEM->init.hwaddr[4] = dev->dev_addr[5];
MEM->init.hwaddr[5] = dev->dev_addr[4];
printk("%s\n", print_mac(mac, dev->dev_addr));
MEM->init.mode = 0x0000;
MEM->init.filter[0] = 0x00000000;
MEM->init.filter[1] = 0x00000000;
MEM->init.rdra = dvma_vtob(MEM->rx_head);
MEM->init.rlen = (RX_LOG_RING_SIZE << 13) |
(dvma_vtob(MEM->rx_head) >> 16);
MEM->init.tdra = dvma_vtob(MEM->tx_head);
MEM->init.tlen = (TX_LOG_RING_SIZE << 13) |
(dvma_vtob(MEM->tx_head) >> 16);
DPRINTK(2, ("initaddr: %08lx rx_ring: %08lx tx_ring: %08lx\n",
dvma_vtob(&(MEM->init)), dvma_vtob(MEM->rx_head),
(dvma_vtob(MEM->tx_head))));
if (did_version++ == 0)
printk( version );
/* The LANCE-specific entries in the device structure. */
dev->open = &lance_open;
dev->hard_start_xmit = &lance_start_xmit;
dev->stop = &lance_close;
dev->set_multicast_list = &set_multicast_list;
dev->set_mac_address = NULL;
// KLUDGE -- REMOVE ME
set_bit(__LINK_STATE_PRESENT, &dev->state);
return 1;
}
static int lance_open( struct net_device *dev )
{
struct lance_private *lp = netdev_priv(dev);
int i;
DPRINTK( 2, ( "%s: lance_open()\n", dev->name ));
REGA(CSR0) = CSR0_STOP;
lance_init_ring(dev);
/* From now on, AREG is kept to point to CSR0 */
REGA(CSR0) = CSR0_INIT;
i = 1000000;
while (--i > 0)
if (DREG & CSR0_IDON)
break;
if (i < 0 || (DREG & CSR0_ERR)) {
DPRINTK( 2, ( "lance_open(): opening %s failed, i=%d, csr0=%04x\n",
dev->name, i, DREG ));
DREG = CSR0_STOP;
return( -EIO );
}
DREG = CSR0_IDON | CSR0_STRT | CSR0_INEA;
netif_start_queue(dev);
DPRINTK( 2, ( "%s: LANCE is open, csr0 %04x\n", dev->name, DREG ));
return( 0 );
}
/* Initialize the LANCE Rx and Tx rings. */
static void lance_init_ring( struct net_device *dev )
{
struct lance_private *lp = netdev_priv(dev);
int i;
lp->lock = 0;
lp->tx_full = 0;
lp->new_rx = lp->new_tx = 0;
lp->old_rx = lp->old_tx = 0;
for( i = 0; i < TX_RING_SIZE; i++ ) {
MEM->tx_head[i].base = dvma_vtob(MEM->tx_data[i]);
MEM->tx_head[i].flag = 0;
MEM->tx_head[i].base_hi =
(dvma_vtob(MEM->tx_data[i])) >>16;
MEM->tx_head[i].length = 0;
MEM->tx_head[i].misc = 0;
}
for( i = 0; i < RX_RING_SIZE; i++ ) {
MEM->rx_head[i].base = dvma_vtob(MEM->rx_data[i]);
MEM->rx_head[i].flag = RMD1_OWN_CHIP;
MEM->rx_head[i].base_hi =
(dvma_vtob(MEM->rx_data[i])) >> 16;
MEM->rx_head[i].buf_length = -PKT_BUF_SZ | 0xf000;
MEM->rx_head[i].msg_length = 0;
}
/* tell the card it's ether address, bytes swapped */
MEM->init.hwaddr[0] = dev->dev_addr[1];
MEM->init.hwaddr[1] = dev->dev_addr[0];
MEM->init.hwaddr[2] = dev->dev_addr[3];
MEM->init.hwaddr[3] = dev->dev_addr[2];
MEM->init.hwaddr[4] = dev->dev_addr[5];
MEM->init.hwaddr[5] = dev->dev_addr[4];
MEM->init.mode = 0x0000;
MEM->init.filter[0] = 0x00000000;
MEM->init.filter[1] = 0x00000000;
MEM->init.rdra = dvma_vtob(MEM->rx_head);
MEM->init.rlen = (RX_LOG_RING_SIZE << 13) |
(dvma_vtob(MEM->rx_head) >> 16);
MEM->init.tdra = dvma_vtob(MEM->tx_head);
MEM->init.tlen = (TX_LOG_RING_SIZE << 13) |
(dvma_vtob(MEM->tx_head) >> 16);
/* tell the lance the address of its init block */
REGA(CSR1) = dvma_vtob(&(MEM->init));
REGA(CSR2) = dvma_vtob(&(MEM->init)) >> 16;
#ifdef CONFIG_SUN3X
REGA(CSR3) = CSR3_BSWP | CSR3_ACON | CSR3_BCON;
#else
REGA(CSR3) = CSR3_BSWP;
#endif
}
static int lance_start_xmit( struct sk_buff *skb, struct net_device *dev )
{
struct lance_private *lp = netdev_priv(dev);
int entry, len;
struct lance_tx_head *head;
unsigned long flags;
DPRINTK( 1, ( "%s: transmit start.\n",
dev->name));
/* Transmitter timeout, serious problems. */
if (netif_queue_stopped(dev)) {
int tickssofar = jiffies - dev->trans_start;
if (tickssofar < 20)
return( 1 );
DPRINTK( 1, ( "%s: transmit timed out, status %04x, resetting.\n",
dev->name, DREG ));
DREG = CSR0_STOP;
/*
* Always set BSWP after a STOP as STOP puts it back into
* little endian mode.
*/
REGA(CSR3) = CSR3_BSWP;
dev->stats.tx_errors++;
if(lance_debug >= 2) {
int i;
printk("Ring data: old_tx %d new_tx %d%s new_rx %d\n",
lp->old_tx, lp->new_tx,
lp->tx_full ? " (full)" : "",
lp->new_rx );
for( i = 0 ; i < RX_RING_SIZE; i++ )
printk( "rx #%d: base=%04x blen=%04x mlen=%04x\n",
i, MEM->rx_head[i].base,
-MEM->rx_head[i].buf_length,
MEM->rx_head[i].msg_length);
for( i = 0 ; i < TX_RING_SIZE; i++ )
printk("tx #%d: base=%04x len=%04x misc=%04x\n",
i, MEM->tx_head[i].base,
-MEM->tx_head[i].length,
MEM->tx_head[i].misc );
}
lance_init_ring(dev);
REGA( CSR0 ) = CSR0_INEA | CSR0_INIT | CSR0_STRT;
netif_start_queue(dev);
dev->trans_start = jiffies;
return 0;
}
/* Block a timer-based transmit from overlapping. This could better be
done with atomic_swap(1, dev->tbusy), but set_bit() works as well. */
/* Block a timer-based transmit from overlapping with us by
stopping the queue for a bit... */
netif_stop_queue(dev);
if (test_and_set_bit( 0, (void*)&lp->lock ) != 0) {
printk( "%s: tx queue lock!.\n", dev->name);
/* don't clear dev->tbusy flag. */
return 1;
}
AREG = CSR0;
DPRINTK( 2, ( "%s: lance_start_xmit() called, csr0 %4.4x.\n",
dev->name, DREG ));
#ifdef CONFIG_SUN3X
/* this weirdness doesn't appear on sun3... */
if(!(DREG & CSR0_INIT)) {
DPRINTK( 1, ("INIT not set, reinitializing...\n"));
REGA( CSR0 ) = CSR0_STOP;
lance_init_ring(dev);
REGA( CSR0 ) = CSR0_INIT | CSR0_STRT;
}
#endif
/* Fill in a Tx ring entry */
#if 0
if (lance_debug >= 2) {
printk( "%s: TX pkt %d type 0x%04x"
" from %s to %s"
" data at 0x%08x len %d\n",
dev->name, lp->new_tx, ((u_short *)skb->data)[6],
DEV_ADDR(&skb->data[6]), DEV_ADDR(skb->data),
(int)skb->data, (int)skb->len );
}
#endif
/* We're not prepared for the int until the last flags are set/reset.
* And the int may happen already after setting the OWN_CHIP... */
local_irq_save(flags);
/* Mask to ring buffer boundary. */
entry = lp->new_tx;
head = &(MEM->tx_head[entry]);
/* Caution: the write order is important here, set the "ownership" bits
* last.
*/
/* the sun3's lance needs it's buffer padded to the minimum
size */
len = (ETH_ZLEN < skb->len) ? skb->len : ETH_ZLEN;
// head->length = -len;
head->length = (-len) | 0xf000;
head->misc = 0;
skb_copy_from_linear_data(skb, PKTBUF_ADDR(head), skb->len);
if (len != skb->len)
memset(PKTBUF_ADDR(head) + skb->len, 0, len-skb->len);
head->flag = TMD1_OWN_CHIP | TMD1_ENP | TMD1_STP;
lp->new_tx = (lp->new_tx + 1) & TX_RING_MOD_MASK;
dev->stats.tx_bytes += skb->len;
/* Trigger an immediate send poll. */
REGA(CSR0) = CSR0_INEA | CSR0_TDMD | CSR0_STRT;
AREG = CSR0;
DPRINTK( 2, ( "%s: lance_start_xmit() exiting, csr0 %4.4x.\n",
dev->name, DREG ));
dev->trans_start = jiffies;
dev_kfree_skb( skb );
lp->lock = 0;
if ((MEM->tx_head[(entry+1) & TX_RING_MOD_MASK].flag & TMD1_OWN) ==
TMD1_OWN_HOST)
netif_start_queue(dev);
local_irq_restore(flags);
return 0;
}
/* The LANCE interrupt handler. */
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
static irqreturn_t lance_interrupt( int irq, void *dev_id)
{
struct net_device *dev = dev_id;
struct lance_private *lp = netdev_priv(dev);
int csr0;
static int in_interrupt;
if (dev == NULL) {
DPRINTK( 1, ( "lance_interrupt(): invalid dev_id\n" ));
return IRQ_NONE;
}
if (in_interrupt)
DPRINTK( 2, ( "%s: Re-entering the interrupt handler.\n", dev->name ));
in_interrupt = 1;
still_more:
flush_cache_all();
AREG = CSR0;
csr0 = DREG;
/* ack interrupts */
DREG = csr0 & (CSR0_TINT | CSR0_RINT | CSR0_IDON);
/* clear errors */
if(csr0 & CSR0_ERR)
DREG = CSR0_BABL | CSR0_MERR | CSR0_CERR | CSR0_MISS;
DPRINTK( 2, ( "%s: interrupt csr0=%04x new csr=%04x.\n",
dev->name, csr0, DREG ));
if (csr0 & CSR0_TINT) { /* Tx-done interrupt */
int old_tx = lp->old_tx;
// if(lance_debug >= 3) {
// int i;
//
// printk("%s: tx int\n", dev->name);
//
// for(i = 0; i < TX_RING_SIZE; i++)
// printk("ring %d flag=%04x\n", i,
// MEM->tx_head[i].flag);
// }
while( old_tx != lp->new_tx) {
struct lance_tx_head *head = &(MEM->tx_head[old_tx]);
DPRINTK(3, ("on tx_ring %d\n", old_tx));
if (head->flag & TMD1_OWN_CHIP)
break; /* It still hasn't been Txed */
if (head->flag & TMD1_ERR) {
int status = head->misc;
dev->stats.tx_errors++;
if (status & TMD3_RTRY) dev->stats.tx_aborted_errors++;
if (status & TMD3_LCAR) dev->stats.tx_carrier_errors++;
if (status & TMD3_LCOL) dev->stats.tx_window_errors++;
if (status & (TMD3_UFLO | TMD3_BUFF)) {
dev->stats.tx_fifo_errors++;
printk("%s: Tx FIFO error\n",
dev->name);
REGA(CSR0) = CSR0_STOP;
REGA(CSR3) = CSR3_BSWP;
lance_init_ring(dev);
REGA(CSR0) = CSR0_STRT | CSR0_INEA;
return IRQ_HANDLED;
}
} else if(head->flag & (TMD1_ENP | TMD1_STP)) {
head->flag &= ~(TMD1_ENP | TMD1_STP);
if(head->flag & (TMD1_ONE | TMD1_MORE))
dev->stats.collisions++;
dev->stats.tx_packets++;
DPRINTK(3, ("cleared tx ring %d\n", old_tx));
}
old_tx = (old_tx +1) & TX_RING_MOD_MASK;
}
lp->old_tx = old_tx;
}
if (netif_queue_stopped(dev)) {
/* The ring is no longer full, clear tbusy. */
netif_start_queue(dev);
netif_wake_queue(dev);
}
if (csr0 & CSR0_RINT) /* Rx interrupt */
lance_rx( dev );
/* Log misc errors. */
if (csr0 & CSR0_BABL) dev->stats.tx_errors++; /* Tx babble. */
if (csr0 & CSR0_MISS) dev->stats.rx_errors++; /* Missed a Rx frame. */
if (csr0 & CSR0_MERR) {
DPRINTK( 1, ( "%s: Bus master arbitration failure (?!?), "
"status %04x.\n", dev->name, csr0 ));
/* Restart the chip. */
REGA(CSR0) = CSR0_STOP;
REGA(CSR3) = CSR3_BSWP;
lance_init_ring(dev);
REGA(CSR0) = CSR0_STRT | CSR0_INEA;
}
/* Clear any other interrupt, and set interrupt enable. */
// DREG = CSR0_BABL | CSR0_CERR | CSR0_MISS | CSR0_MERR |
// CSR0_IDON | CSR0_INEA;
REGA(CSR0) = CSR0_INEA;
if(DREG & (CSR0_RINT | CSR0_TINT)) {
DPRINTK(2, ("restarting interrupt, csr0=%#04x\n", DREG));
goto still_more;
}
DPRINTK( 2, ( "%s: exiting interrupt, csr0=%#04x.\n",
dev->name, DREG ));
in_interrupt = 0;
return IRQ_HANDLED;
}
/* get packet, toss into skbuff */
static int lance_rx( struct net_device *dev )
{
struct lance_private *lp = netdev_priv(dev);
int entry = lp->new_rx;
/* If we own the next entry, it's a new packet. Send it up. */
while( (MEM->rx_head[entry].flag & RMD1_OWN) == RMD1_OWN_HOST ) {
struct lance_rx_head *head = &(MEM->rx_head[entry]);
int status = head->flag;
if (status != (RMD1_ENP|RMD1_STP)) { /* There was an error. */
/* There is a tricky error noted by John Murphy,
<murf@perftech.com> to Russ Nelson: Even with
full-sized buffers it's possible for a jabber packet to use two
buffers, with only the last correctly noting the error. */
if (status & RMD1_ENP) /* Only count a general error at the */
dev->stats.rx_errors++; /* end of a packet.*/
if (status & RMD1_FRAM) dev->stats.rx_frame_errors++;
if (status & RMD1_OFLO) dev->stats.rx_over_errors++;
if (status & RMD1_CRC) dev->stats.rx_crc_errors++;
if (status & RMD1_BUFF) dev->stats.rx_fifo_errors++;
head->flag &= (RMD1_ENP|RMD1_STP);
} else {
/* Malloc up new buffer, compatible with net-3. */
// short pkt_len = head->msg_length;// & 0xfff;
short pkt_len = (head->msg_length & 0xfff) - 4;
struct sk_buff *skb;
if (pkt_len < 60) {
printk( "%s: Runt packet!\n", dev->name );
dev->stats.rx_errors++;
}
else {
skb = dev_alloc_skb( pkt_len+2 );
if (skb == NULL) {
DPRINTK( 1, ( "%s: Memory squeeze, deferring packet.\n",
dev->name ));
dev->stats.rx_dropped++;
head->msg_length = 0;
head->flag |= RMD1_OWN_CHIP;
lp->new_rx = (lp->new_rx+1) &
RX_RING_MOD_MASK;
}
#if 0
if (lance_debug >= 3) {
u_char *data = PKTBUF_ADDR(head);
DECLARE_MAC_BUF(mac);
DECLARE_MAC_BUF(mac2)
printk("%s: RX pkt %d type 0x%04x"
" from %s to %s",
dev->name, lp->new_tx, ((u_short *)data)[6],
print_mac(mac, &data[6]), print_mac(mac2, data));
printk(" data %02x %02x %02x %02x %02x %02x %02x %02x "
"len %d at %08x\n",
data[15], data[16], data[17], data[18],
data[19], data[20], data[21], data[22],
pkt_len, data);
}
#endif
if (lance_debug >= 3) {
u_char *data = PKTBUF_ADDR(head);
printk( "%s: RX pkt %d type 0x%04x len %d\n ", dev->name, entry, ((u_short *)data)[6], pkt_len);
}
skb_reserve( skb, 2 ); /* 16 byte align */
skb_put( skb, pkt_len ); /* Make room */
skb_copy_to_linear_data(skb,
PKTBUF_ADDR(head),
pkt_len);
skb->protocol = eth_type_trans( skb, dev );
netif_rx( skb );
dev->last_rx = jiffies;
dev->stats.rx_packets++;
dev->stats.rx_bytes += pkt_len;
}
}
// head->buf_length = -PKT_BUF_SZ | 0xf000;
head->msg_length = 0;
head->flag = RMD1_OWN_CHIP;
entry = lp->new_rx = (lp->new_rx +1) & RX_RING_MOD_MASK;
}
/* From lance.c (Donald Becker): */
/* We should check that at least two ring entries are free.
If not, we should free one and mark stats->rx_dropped++. */
return 0;
}
static int lance_close( struct net_device *dev )
{
struct lance_private *lp = netdev_priv(dev);
netif_stop_queue(dev);
AREG = CSR0;
DPRINTK( 2, ( "%s: Shutting down ethercard, status was %2.2x.\n",
dev->name, DREG ));
/* We stop the LANCE here -- it occasionally polls
memory if we don't. */
DREG = CSR0_STOP;
return 0;
}
/* Set or clear the multicast filter for this adaptor.
num_addrs == -1 Promiscuous mode, receive all packets
num_addrs == 0 Normal mode, clear multicast list
num_addrs > 0 Multicast mode, receive normal and MC packets, and do
best-effort filtering.
*/
/* completely untested on a sun3 */
static void set_multicast_list( struct net_device *dev )
{
struct lance_private *lp = netdev_priv(dev);
if(netif_queue_stopped(dev))
/* Only possible if board is already started */
return;
/* We take the simple way out and always enable promiscuous mode. */
DREG = CSR0_STOP; /* Temporarily stop the lance. */
if (dev->flags & IFF_PROMISC) {
/* Log any net taps. */
DPRINTK( 3, ( "%s: Promiscuous mode enabled.\n", dev->name ));
REGA( CSR15 ) = 0x8000; /* Set promiscuous mode */
} else {
short multicast_table[4];
int num_addrs = dev->mc_count;
int i;
/* We don't use the multicast table, but rely on upper-layer
* filtering. */
memset( multicast_table, (num_addrs == 0) ? 0 : -1,
sizeof(multicast_table) );
for( i = 0; i < 4; i++ )
REGA( CSR8+i ) = multicast_table[i];
REGA( CSR15 ) = 0; /* Unset promiscuous mode */
}
/*
* Always set BSWP after a STOP as STOP puts it back into
* little endian mode.
*/
REGA( CSR3 ) = CSR3_BSWP;
/* Resume normal operation and reset AREG to CSR0 */
REGA( CSR0 ) = CSR0_IDON | CSR0_INEA | CSR0_STRT;
}
#ifdef MODULE
static struct net_device *sun3lance_dev;
int __init init_module(void)
{
sun3lance_dev = sun3lance_probe(-1);
if (IS_ERR(sun3lance_dev))
return PTR_ERR(sun3lance_dev);
return 0;
}
void __exit cleanup_module(void)
{
unregister_netdev(sun3lance_dev);
#ifdef CONFIG_SUN3
iounmap((void __iomem *)sun3lance_dev->base_addr);
#endif
free_netdev(sun3lance_dev);
}
#endif /* MODULE */