dect
/
linux-2.6
Archived
13
0
Fork 0
This repository has been archived on 2022-02-17. You can view files and clone it, but cannot push or open issues or pull requests.
linux-2.6/drivers/net/macmace.c

711 lines
16 KiB
C
Raw Normal View History

/*
* Driver for the Macintosh 68K onboard MACE controller with PSC
* driven DMA. The MACE driver code is derived from mace.c. The
* Mac68k theory of operation is courtesy of the MacBSD wizards.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*
* Copyright (C) 1996 Paul Mackerras.
* Copyright (C) 1998 Alan Cox <alan@redhat.com>
*
* Modified heavily by Joshua M. Thompson based on Dave Huang's NetBSD driver
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/delay.h>
#include <linux/string.h>
#include <linux/crc32.h>
#include <asm/io.h>
#include <asm/pgtable.h>
#include <asm/irq.h>
#include <asm/macintosh.h>
#include <asm/macints.h>
#include <asm/mac_psc.h>
#include <asm/page.h>
#include "mace.h"
#define N_TX_RING 1
#define N_RX_RING 8
#define N_RX_PAGES ((N_RX_RING * 0x0800 + PAGE_SIZE - 1) / PAGE_SIZE)
#define TX_TIMEOUT HZ
/* Bits in transmit DMA status */
#define TX_DMA_ERR 0x80
/* The MACE is simply wired down on a Mac68K box */
#define MACE_BASE (void *)(0x50F1C000)
#define MACE_PROM (void *)(0x50F08001)
struct mace_data {
volatile struct mace *mace;
volatile unsigned char *tx_ring;
volatile unsigned char *tx_ring_phys;
volatile unsigned char *rx_ring;
volatile unsigned char *rx_ring_phys;
int dma_intr;
struct net_device_stats stats;
int rx_slot, rx_tail;
int tx_slot, tx_sloti, tx_count;
};
struct mace_frame {
u16 len;
u16 status;
u16 rntpc;
u16 rcvcc;
u32 pad1;
u32 pad2;
u8 data[1];
/* And frame continues.. */
};
#define PRIV_BYTES sizeof(struct mace_data)
extern void psc_debug_dump(void);
static int mace_open(struct net_device *dev);
static int mace_close(struct net_device *dev);
static int mace_xmit_start(struct sk_buff *skb, struct net_device *dev);
static struct net_device_stats *mace_stats(struct net_device *dev);
static void mace_set_multicast(struct net_device *dev);
static int mace_set_address(struct net_device *dev, void *addr);
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 mace_interrupt(int irq, void *dev_id);
static irqreturn_t mace_dma_intr(int irq, void *dev_id);
static void mace_tx_timeout(struct net_device *dev);
/* Bit-reverse one byte of an ethernet hardware address. */
static int bitrev(int b)
{
int d = 0, i;
for (i = 0; i < 8; ++i, b >>= 1) {
d = (d << 1) | (b & 1);
}
return d;
}
/*
* Load a receive DMA channel with a base address and ring length
*/
static void mace_load_rxdma_base(struct net_device *dev, int set)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
psc_write_word(PSC_ENETRD_CMD + set, 0x0100);
psc_write_long(PSC_ENETRD_ADDR + set, (u32) mp->rx_ring_phys);
psc_write_long(PSC_ENETRD_LEN + set, N_RX_RING);
psc_write_word(PSC_ENETRD_CMD + set, 0x9800);
mp->rx_tail = 0;
}
/*
* Reset the receive DMA subsystem
*/
static void mace_rxdma_reset(struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mace = mp->mace;
u8 maccc = mace->maccc;
mace->maccc = maccc & ~ENRCV;
psc_write_word(PSC_ENETRD_CTL, 0x8800);
mace_load_rxdma_base(dev, 0x00);
psc_write_word(PSC_ENETRD_CTL, 0x0400);
psc_write_word(PSC_ENETRD_CTL, 0x8800);
mace_load_rxdma_base(dev, 0x10);
psc_write_word(PSC_ENETRD_CTL, 0x0400);
mace->maccc = maccc;
mp->rx_slot = 0;
psc_write_word(PSC_ENETRD_CMD + PSC_SET0, 0x9800);
psc_write_word(PSC_ENETRD_CMD + PSC_SET1, 0x9800);
}
/*
* Reset the transmit DMA subsystem
*/
static void mace_txdma_reset(struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mace = mp->mace;
u8 maccc;
psc_write_word(PSC_ENETWR_CTL, 0x8800);
maccc = mace->maccc;
mace->maccc = maccc & ~ENXMT;
mp->tx_slot = mp->tx_sloti = 0;
mp->tx_count = N_TX_RING;
psc_write_word(PSC_ENETWR_CTL, 0x0400);
mace->maccc = maccc;
}
/*
* Disable DMA
*/
static void mace_dma_off(struct net_device *dev)
{
psc_write_word(PSC_ENETRD_CTL, 0x8800);
psc_write_word(PSC_ENETRD_CTL, 0x1000);
psc_write_word(PSC_ENETRD_CMD + PSC_SET0, 0x1100);
psc_write_word(PSC_ENETRD_CMD + PSC_SET1, 0x1100);
psc_write_word(PSC_ENETWR_CTL, 0x8800);
psc_write_word(PSC_ENETWR_CTL, 0x1000);
psc_write_word(PSC_ENETWR_CMD + PSC_SET0, 0x1100);
psc_write_word(PSC_ENETWR_CMD + PSC_SET1, 0x1100);
}
/*
* Not really much of a probe. The hardware table tells us if this
* model of Macintrash has a MACE (AV macintoshes)
*/
struct net_device *mace_probe(int unit)
{
int j;
struct mace_data *mp;
unsigned char *addr;
struct net_device *dev;
unsigned char checksum = 0;
static int found = 0;
int err;
if (found || macintosh_config->ether_type != MAC_ETHER_MACE)
return ERR_PTR(-ENODEV);
found = 1; /* prevent 'finding' one on every device probe */
dev = alloc_etherdev(PRIV_BYTES);
if (!dev)
return ERR_PTR(-ENOMEM);
if (unit >= 0)
sprintf(dev->name, "eth%d", unit);
mp = (struct mace_data *) dev->priv;
dev->base_addr = (u32)MACE_BASE;
mp->mace = (volatile struct mace *) MACE_BASE;
dev->irq = IRQ_MAC_MACE;
mp->dma_intr = IRQ_MAC_MACE_DMA;
/*
* The PROM contains 8 bytes which total 0xFF when XOR'd
* together. Due to the usual peculiar apple brain damage
* the bytes are spaced out in a strange boundary and the
* bits are reversed.
*/
addr = (void *)MACE_PROM;
for (j = 0; j < 6; ++j) {
u8 v=bitrev(addr[j<<4]);
checksum ^= v;
dev->dev_addr[j] = v;
}
for (; j < 8; ++j) {
checksum ^= bitrev(addr[j<<4]);
}
if (checksum != 0xFF) {
free_netdev(dev);
return ERR_PTR(-ENODEV);
}
memset(&mp->stats, 0, sizeof(mp->stats));
dev->open = mace_open;
dev->stop = mace_close;
dev->hard_start_xmit = mace_xmit_start;
dev->tx_timeout = mace_tx_timeout;
dev->watchdog_timeo = TX_TIMEOUT;
dev->get_stats = mace_stats;
dev->set_multicast_list = mace_set_multicast;
dev->set_mac_address = mace_set_address;
printk(KERN_INFO "%s: 68K MACE, hardware address %.2X", dev->name, dev->dev_addr[0]);
for (j = 1 ; j < 6 ; j++) printk(":%.2X", dev->dev_addr[j]);
printk("\n");
err = register_netdev(dev);
if (!err)
return dev;
free_netdev(dev);
return ERR_PTR(err);
}
/*
* Load the address on a mace controller.
*/
static int mace_set_address(struct net_device *dev, void *addr)
{
unsigned char *p = addr;
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mb = mp->mace;
int i;
unsigned long flags;
u8 maccc;
local_irq_save(flags);
maccc = mb->maccc;
/* load up the hardware address */
mb->iac = ADDRCHG | PHYADDR;
while ((mb->iac & ADDRCHG) != 0);
for (i = 0; i < 6; ++i) {
mb->padr = dev->dev_addr[i] = p[i];
}
mb->maccc = maccc;
local_irq_restore(flags);
return 0;
}
/*
* Open the Macintosh MACE. Most of this is playing with the DMA
* engine. The ethernet chip is quite friendly.
*/
static int mace_open(struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mb = mp->mace;
#if 0
int i;
i = 200;
while (--i) {
mb->biucc = SWRST;
if (mb->biucc & SWRST) {
udelay(10);
continue;
}
break;
}
if (!i) {
printk(KERN_ERR "%s: software reset failed!!\n", dev->name);
return -EAGAIN;
}
#endif
mb->biucc = XMTSP_64;
mb->fifocc = XMTFW_16 | RCVFW_64 | XMTFWU | RCVFWU | XMTBRST | RCVBRST;
mb->xmtfc = AUTO_PAD_XMIT;
mb->plscc = PORTSEL_AUI;
/* mb->utr = RTRD; */
if (request_irq(dev->irq, mace_interrupt, 0, dev->name, dev)) {
printk(KERN_ERR "%s: can't get irq %d\n", dev->name, dev->irq);
return -EAGAIN;
}
if (request_irq(mp->dma_intr, mace_dma_intr, 0, dev->name, dev)) {
printk(KERN_ERR "%s: can't get irq %d\n", dev->name, mp->dma_intr);
free_irq(dev->irq, dev);
return -EAGAIN;
}
/* Allocate the DMA ring buffers */
mp->rx_ring = (void *) __get_free_pages(GFP_KERNEL | GFP_DMA, N_RX_PAGES);
mp->tx_ring = (void *) __get_free_pages(GFP_KERNEL | GFP_DMA, 0);
if (mp->tx_ring==NULL || mp->rx_ring==NULL) {
if (mp->rx_ring) free_pages((u32) mp->rx_ring, N_RX_PAGES);
if (mp->tx_ring) free_pages((u32) mp->tx_ring, 0);
free_irq(dev->irq, dev);
free_irq(mp->dma_intr, dev);
printk(KERN_ERR "%s: unable to allocate DMA buffers\n", dev->name);
return -ENOMEM;
}
mp->rx_ring_phys = (unsigned char *) virt_to_bus((void *)mp->rx_ring);
mp->tx_ring_phys = (unsigned char *) virt_to_bus((void *)mp->tx_ring);
/* We want the Rx buffer to be uncached and the Tx buffer to be writethrough */
kernel_set_cachemode((void *)mp->rx_ring, N_RX_PAGES * PAGE_SIZE, IOMAP_NOCACHE_NONSER);
kernel_set_cachemode((void *)mp->tx_ring, PAGE_SIZE, IOMAP_WRITETHROUGH);
mace_dma_off(dev);
/* Not sure what these do */
psc_write_word(PSC_ENETWR_CTL, 0x9000);
psc_write_word(PSC_ENETRD_CTL, 0x9000);
psc_write_word(PSC_ENETWR_CTL, 0x0400);
psc_write_word(PSC_ENETRD_CTL, 0x0400);
#if 0
/* load up the hardware address */
mb->iac = ADDRCHG | PHYADDR;
while ((mb->iac & ADDRCHG) != 0);
for (i = 0; i < 6; ++i)
mb->padr = dev->dev_addr[i];
/* clear the multicast filter */
mb->iac = ADDRCHG | LOGADDR;
while ((mb->iac & ADDRCHG) != 0);
for (i = 0; i < 8; ++i)
mb->ladrf = 0;
mb->plscc = PORTSEL_GPSI + ENPLSIO;
mb->maccc = ENXMT | ENRCV;
mb->imr = RCVINT;
#endif
mace_rxdma_reset(dev);
mace_txdma_reset(dev);
return 0;
}
/*
* Shut down the mace and its interrupt channel
*/
static int mace_close(struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mb = mp->mace;
mb->maccc = 0; /* disable rx and tx */
mb->imr = 0xFF; /* disable all irqs */
mace_dma_off(dev); /* disable rx and tx dma */
free_irq(dev->irq, dev);
free_irq(IRQ_MAC_MACE_DMA, dev);
free_pages((u32) mp->rx_ring, N_RX_PAGES);
free_pages((u32) mp->tx_ring, 0);
return 0;
}
/*
* Transmit a frame
*/
static int mace_xmit_start(struct sk_buff *skb, struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
/* Stop the queue if the buffer is full */
if (!mp->tx_count) {
netif_stop_queue(dev);
return 1;
}
mp->tx_count--;
mp->stats.tx_packets++;
mp->stats.tx_bytes += skb->len;
/* We need to copy into our xmit buffer to take care of alignment and caching issues */
memcpy((void *) mp->tx_ring, skb->data, skb->len);
/* load the Tx DMA and fire it off */
psc_write_long(PSC_ENETWR_ADDR + mp->tx_slot, (u32) mp->tx_ring_phys);
psc_write_long(PSC_ENETWR_LEN + mp->tx_slot, skb->len);
psc_write_word(PSC_ENETWR_CMD + mp->tx_slot, 0x9800);
mp->tx_slot ^= 0x10;
dev_kfree_skb(skb);
return 0;
}
static struct net_device_stats *mace_stats(struct net_device *dev)
{
struct mace_data *p = (struct mace_data *) dev->priv;
return &p->stats;
}
static void mace_set_multicast(struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mb = mp->mace;
int i, j;
u32 crc;
u8 maccc;
maccc = mb->maccc;
mb->maccc &= ~PROM;
if (dev->flags & IFF_PROMISC) {
mb->maccc |= PROM;
} else {
unsigned char multicast_filter[8];
struct dev_mc_list *dmi = dev->mc_list;
if (dev->flags & IFF_ALLMULTI) {
for (i = 0; i < 8; i++) {
multicast_filter[i] = 0xFF;
}
} else {
for (i = 0; i < 8; i++)
multicast_filter[i] = 0;
for (i = 0; i < dev->mc_count; i++) {
crc = ether_crc_le(6, dmi->dmi_addr);
j = crc >> 26; /* bit number in multicast_filter */
multicast_filter[j >> 3] |= 1 << (j & 7);
dmi = dmi->next;
}
}
mb->iac = ADDRCHG | LOGADDR;
while (mb->iac & ADDRCHG);
for (i = 0; i < 8; ++i) {
mb->ladrf = multicast_filter[i];
}
}
mb->maccc = maccc;
}
/*
* Miscellaneous interrupts are handled here. We may end up
* having to bash the chip on the head for bad errors
*/
static void mace_handle_misc_intrs(struct mace_data *mp, int intr)
{
volatile struct mace *mb = mp->mace;
static int mace_babbles, mace_jabbers;
if (intr & MPCO) {
mp->stats.rx_missed_errors += 256;
}
mp->stats.rx_missed_errors += mb->mpc; /* reading clears it */
if (intr & RNTPCO) {
mp->stats.rx_length_errors += 256;
}
mp->stats.rx_length_errors += mb->rntpc; /* reading clears it */
if (intr & CERR) {
++mp->stats.tx_heartbeat_errors;
}
if (intr & BABBLE) {
if (mace_babbles++ < 4) {
printk(KERN_DEBUG "mace: babbling transmitter\n");
}
}
if (intr & JABBER) {
if (mace_jabbers++ < 4) {
printk(KERN_DEBUG "mace: jabbering transceiver\n");
}
}
}
/*
* A transmit error has occurred. (We kick the transmit side from
* the DMA completion)
*/
static void mace_xmit_error(struct net_device *dev)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mb = mp->mace;
u8 xmtfs, xmtrc;
xmtfs = mb->xmtfs;
xmtrc = mb->xmtrc;
if (xmtfs & XMTSV) {
if (xmtfs & UFLO) {
printk("%s: DMA underrun.\n", dev->name);
mp->stats.tx_errors++;
mp->stats.tx_fifo_errors++;
mace_txdma_reset(dev);
}
if (xmtfs & RTRY) {
mp->stats.collisions++;
}
}
}
/*
* A receive interrupt occurred.
*/
static void mace_recv_interrupt(struct net_device *dev)
{
/* struct mace_data *mp = (struct mace_data *) dev->priv; */
// volatile struct mace *mb = mp->mace;
}
/*
* Process the chip interrupt
*/
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 mace_interrupt(int irq, void *dev_id)
{
struct net_device *dev = (struct net_device *) dev_id;
struct mace_data *mp = (struct mace_data *) dev->priv;
volatile struct mace *mb = mp->mace;
u8 ir;
ir = mb->ir;
mace_handle_misc_intrs(mp, ir);
if (ir & XMTINT) {
mace_xmit_error(dev);
}
if (ir & RCVINT) {
mace_recv_interrupt(dev);
}
return IRQ_HANDLED;
}
static void mace_tx_timeout(struct net_device *dev)
{
/* struct mace_data *mp = (struct mace_data *) dev->priv; */
// volatile struct mace *mb = mp->mace;
}
/*
* Handle a newly arrived frame
*/
static void mace_dma_rx_frame(struct net_device *dev, struct mace_frame *mf)
{
struct mace_data *mp = (struct mace_data *) dev->priv;
struct sk_buff *skb;
if (mf->status & RS_OFLO) {
printk("%s: fifo overflow.\n", dev->name);
mp->stats.rx_errors++;
mp->stats.rx_fifo_errors++;
}
if (mf->status&(RS_CLSN|RS_FRAMERR|RS_FCSERR))
mp->stats.rx_errors++;
if (mf->status&RS_CLSN) {
mp->stats.collisions++;
}
if (mf->status&RS_FRAMERR) {
mp->stats.rx_frame_errors++;
}
if (mf->status&RS_FCSERR) {
mp->stats.rx_crc_errors++;
}
skb = dev_alloc_skb(mf->len+2);
if (!skb) {
mp->stats.rx_dropped++;
return;
}
skb_reserve(skb,2);
memcpy(skb_put(skb, mf->len), mf->data, mf->len);
skb->dev = dev;
skb->protocol = eth_type_trans(skb, dev);
netif_rx(skb);
dev->last_rx = jiffies;
mp->stats.rx_packets++;
mp->stats.rx_bytes += mf->len;
}
/*
* The PSC has passed us a DMA interrupt event.
*/
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 mace_dma_intr(int irq, void *dev_id)
{
struct net_device *dev = (struct net_device *) dev_id;
struct mace_data *mp = (struct mace_data *) dev->priv;
int left, head;
u16 status;
u32 baka;
/* Not sure what this does */
while ((baka = psc_read_long(PSC_MYSTERY)) != psc_read_long(PSC_MYSTERY));
if (!(baka & 0x60000000)) return IRQ_NONE;
/*
* Process the read queue
*/
status = psc_read_word(PSC_ENETRD_CTL);
if (status & 0x2000) {
mace_rxdma_reset(dev);
} else if (status & 0x0100) {
psc_write_word(PSC_ENETRD_CMD + mp->rx_slot, 0x1100);
left = psc_read_long(PSC_ENETRD_LEN + mp->rx_slot);
head = N_RX_RING - left;
/* Loop through the ring buffer and process new packages */
while (mp->rx_tail < head) {
mace_dma_rx_frame(dev, (struct mace_frame *) (mp->rx_ring + (mp->rx_tail * 0x0800)));
mp->rx_tail++;
}
/* If we're out of buffers in this ring then switch to */
/* the other set, otherwise just reactivate this one. */
if (!left) {
mace_load_rxdma_base(dev, mp->rx_slot);
mp->rx_slot ^= 0x10;
} else {
psc_write_word(PSC_ENETRD_CMD + mp->rx_slot, 0x9800);
}
}
/*
* Process the write queue
*/
status = psc_read_word(PSC_ENETWR_CTL);
if (status & 0x2000) {
mace_txdma_reset(dev);
} else if (status & 0x0100) {
psc_write_word(PSC_ENETWR_CMD + mp->tx_sloti, 0x0100);
mp->tx_sloti ^= 0x10;
mp->tx_count++;
netif_wake_queue(dev);
}
return IRQ_HANDLED;
}
MODULE_LICENSE("GPL");