linux/drivers/net/ethernet/apple/bmac.c

1683 lines
41 KiB
C
Raw Normal View History

/*
* Network device driver for the BMAC ethernet controller on
* Apple Powermacs. Assumes it's under a DBDMA controller.
*
* Copyright (C) 1998 Randy Gobbel.
*
* May 1999, Al Viro: proper release of /proc/net/bmac entry, switched to
* dynamic procfs inode.
*/
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/delay.h>
#include <linux/string.h>
#include <linux/timer.h>
#include <linux/proc_fs.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/crc32.h>
#include <linux/bitrev.h>
#include <linux/ethtool.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
#include <linux/slab.h>
#include <asm/prom.h>
#include <asm/dbdma.h>
#include <asm/io.h>
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/machdep.h>
#include <asm/pmac_feature.h>
#include <asm/macio.h>
#include <asm/irq.h>
#include "bmac.h"
#define trunc_page(x) ((void *)(((unsigned long)(x)) & ~((unsigned long)(PAGE_SIZE - 1))))
#define round_page(x) trunc_page(((unsigned long)(x)) + ((unsigned long)(PAGE_SIZE - 1)))
/*
* CRC polynomial - used in working out multicast filter bits.
*/
#define ENET_CRCPOLY 0x04c11db7
/* switch to use multicast code lifted from sunhme driver */
#define SUNHME_MULTICAST
#define N_RX_RING 64
#define N_TX_RING 32
#define MAX_TX_ACTIVE 1
#define ETHERCRC 4
#define ETHERMINPACKET 64
#define ETHERMTU 1500
#define RX_BUFLEN (ETHERMTU + 14 + ETHERCRC + 2)
#define TX_TIMEOUT HZ /* 1 second */
/* Bits in transmit DMA status */
#define TX_DMA_ERR 0x80
#define XXDEBUG(args)
struct bmac_data {
/* volatile struct bmac *bmac; */
struct sk_buff_head *queue;
volatile struct dbdma_regs __iomem *tx_dma;
int tx_dma_intr;
volatile struct dbdma_regs __iomem *rx_dma;
int rx_dma_intr;
volatile struct dbdma_cmd *tx_cmds; /* xmit dma command list */
volatile struct dbdma_cmd *rx_cmds; /* recv dma command list */
struct macio_dev *mdev;
int is_bmac_plus;
struct sk_buff *rx_bufs[N_RX_RING];
int rx_fill;
int rx_empty;
struct sk_buff *tx_bufs[N_TX_RING];
int tx_fill;
int tx_empty;
unsigned char tx_fullup;
struct timer_list tx_timeout;
int timeout_active;
int sleeping;
int opened;
unsigned short hash_use_count[64];
unsigned short hash_table_mask[4];
spinlock_t lock;
};
#if 0 /* Move that to ethtool */
typedef struct bmac_reg_entry {
char *name;
unsigned short reg_offset;
} bmac_reg_entry_t;
#define N_REG_ENTRIES 31
static bmac_reg_entry_t reg_entries[N_REG_ENTRIES] = {
{"MEMADD", MEMADD},
{"MEMDATAHI", MEMDATAHI},
{"MEMDATALO", MEMDATALO},
{"TXPNTR", TXPNTR},
{"RXPNTR", RXPNTR},
{"IPG1", IPG1},
{"IPG2", IPG2},
{"ALIMIT", ALIMIT},
{"SLOT", SLOT},
{"PALEN", PALEN},
{"PAPAT", PAPAT},
{"TXSFD", TXSFD},
{"JAM", JAM},
{"TXCFG", TXCFG},
{"TXMAX", TXMAX},
{"TXMIN", TXMIN},
{"PAREG", PAREG},
{"DCNT", DCNT},
{"NCCNT", NCCNT},
{"NTCNT", NTCNT},
{"EXCNT", EXCNT},
{"LTCNT", LTCNT},
{"TXSM", TXSM},
{"RXCFG", RXCFG},
{"RXMAX", RXMAX},
{"RXMIN", RXMIN},
{"FRCNT", FRCNT},
{"AECNT", AECNT},
{"FECNT", FECNT},
{"RXSM", RXSM},
{"RXCV", RXCV}
};
#endif
static unsigned char *bmac_emergency_rxbuf;
/*
* Number of bytes of private data per BMAC: allow enough for
* the rx and tx dma commands plus a branch dma command each,
* and another 16 bytes to allow us to align the dma command
* buffers on a 16 byte boundary.
*/
#define PRIV_BYTES (sizeof(struct bmac_data) \
+ (N_RX_RING + N_TX_RING + 4) * sizeof(struct dbdma_cmd) \
+ sizeof(struct sk_buff_head))
static int bmac_open(struct net_device *dev);
static int bmac_close(struct net_device *dev);
static int bmac_transmit_packet(struct sk_buff *skb, struct net_device *dev);
static void bmac_set_multicast(struct net_device *dev);
static void bmac_reset_and_enable(struct net_device *dev);
static void bmac_start_chip(struct net_device *dev);
static void bmac_init_chip(struct net_device *dev);
static void bmac_init_registers(struct net_device *dev);
static void bmac_enable_and_reset_chip(struct net_device *dev);
static int bmac_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 bmac_misc_intr(int irq, void *dev_id);
static irqreturn_t bmac_txdma_intr(int irq, void *dev_id);
static irqreturn_t bmac_rxdma_intr(int irq, void *dev_id);
static void bmac_set_timeout(struct net_device *dev);
static void bmac_tx_timeout(unsigned long data);
static int bmac_output(struct sk_buff *skb, struct net_device *dev);
static void bmac_start(struct net_device *dev);
#define DBDMA_SET(x) ( ((x) | (x) << 16) )
#define DBDMA_CLEAR(x) ( (x) << 16)
static inline void
dbdma_st32(volatile __u32 __iomem *a, unsigned long x)
{
__asm__ volatile( "stwbrx %0,0,%1" : : "r" (x), "r" (a) : "memory");
}
static inline unsigned long
dbdma_ld32(volatile __u32 __iomem *a)
{
__u32 swap;
__asm__ volatile ("lwbrx %0,0,%1" : "=r" (swap) : "r" (a));
return swap;
}
static void
dbdma_continue(volatile struct dbdma_regs __iomem *dmap)
{
dbdma_st32(&dmap->control,
DBDMA_SET(RUN|WAKE) | DBDMA_CLEAR(PAUSE|DEAD));
eieio();
}
static void
dbdma_reset(volatile struct dbdma_regs __iomem *dmap)
{
dbdma_st32(&dmap->control,
DBDMA_CLEAR(ACTIVE|DEAD|WAKE|FLUSH|PAUSE|RUN));
eieio();
while (dbdma_ld32(&dmap->status) & RUN)
eieio();
}
static void
dbdma_setcmd(volatile struct dbdma_cmd *cp,
unsigned short cmd, unsigned count, unsigned long addr,
unsigned long cmd_dep)
{
out_le16(&cp->command, cmd);
out_le16(&cp->req_count, count);
out_le32(&cp->phy_addr, addr);
out_le32(&cp->cmd_dep, cmd_dep);
out_le16(&cp->xfer_status, 0);
out_le16(&cp->res_count, 0);
}
static inline
void bmwrite(struct net_device *dev, unsigned long reg_offset, unsigned data )
{
out_le16((void __iomem *)dev->base_addr + reg_offset, data);
}
static inline
unsigned short bmread(struct net_device *dev, unsigned long reg_offset )
{
return in_le16((void __iomem *)dev->base_addr + reg_offset);
}
static void
bmac_enable_and_reset_chip(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
volatile struct dbdma_regs __iomem *td = bp->tx_dma;
if (rd)
dbdma_reset(rd);
if (td)
dbdma_reset(td);
pmac_call_feature(PMAC_FTR_BMAC_ENABLE, macio_get_of_node(bp->mdev), 0, 1);
}
#define MIFDELAY udelay(10)
static unsigned int
bmac_mif_readbits(struct net_device *dev, int nb)
{
unsigned int val = 0;
while (--nb >= 0) {
bmwrite(dev, MIFCSR, 0);
MIFDELAY;
if (bmread(dev, MIFCSR) & 8)
val |= 1 << nb;
bmwrite(dev, MIFCSR, 1);
MIFDELAY;
}
bmwrite(dev, MIFCSR, 0);
MIFDELAY;
bmwrite(dev, MIFCSR, 1);
MIFDELAY;
return val;
}
static void
bmac_mif_writebits(struct net_device *dev, unsigned int val, int nb)
{
int b;
while (--nb >= 0) {
b = (val & (1 << nb))? 6: 4;
bmwrite(dev, MIFCSR, b);
MIFDELAY;
bmwrite(dev, MIFCSR, b|1);
MIFDELAY;
}
}
static unsigned int
bmac_mif_read(struct net_device *dev, unsigned int addr)
{
unsigned int val;
bmwrite(dev, MIFCSR, 4);
MIFDELAY;
bmac_mif_writebits(dev, ~0U, 32);
bmac_mif_writebits(dev, 6, 4);
bmac_mif_writebits(dev, addr, 10);
bmwrite(dev, MIFCSR, 2);
MIFDELAY;
bmwrite(dev, MIFCSR, 1);
MIFDELAY;
val = bmac_mif_readbits(dev, 17);
bmwrite(dev, MIFCSR, 4);
MIFDELAY;
return val;
}
static void
bmac_mif_write(struct net_device *dev, unsigned int addr, unsigned int val)
{
bmwrite(dev, MIFCSR, 4);
MIFDELAY;
bmac_mif_writebits(dev, ~0U, 32);
bmac_mif_writebits(dev, 5, 4);
bmac_mif_writebits(dev, addr, 10);
bmac_mif_writebits(dev, 2, 2);
bmac_mif_writebits(dev, val, 16);
bmac_mif_writebits(dev, 3, 2);
}
static void
bmac_init_registers(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
volatile unsigned short regValue;
unsigned short *pWord16;
int i;
/* XXDEBUG(("bmac: enter init_registers\n")); */
bmwrite(dev, RXRST, RxResetValue);
bmwrite(dev, TXRST, TxResetBit);
i = 100;
do {
--i;
udelay(10000);
regValue = bmread(dev, TXRST); /* wait for reset to clear..acknowledge */
} while ((regValue & TxResetBit) && i > 0);
if (!bp->is_bmac_plus) {
regValue = bmread(dev, XCVRIF);
regValue |= ClkBit | SerialMode | COLActiveLow;
bmwrite(dev, XCVRIF, regValue);
udelay(10000);
}
bmwrite(dev, RSEED, (unsigned short)0x1968);
regValue = bmread(dev, XIFC);
regValue |= TxOutputEnable;
bmwrite(dev, XIFC, regValue);
bmread(dev, PAREG);
/* set collision counters to 0 */
bmwrite(dev, NCCNT, 0);
bmwrite(dev, NTCNT, 0);
bmwrite(dev, EXCNT, 0);
bmwrite(dev, LTCNT, 0);
/* set rx counters to 0 */
bmwrite(dev, FRCNT, 0);
bmwrite(dev, LECNT, 0);
bmwrite(dev, AECNT, 0);
bmwrite(dev, FECNT, 0);
bmwrite(dev, RXCV, 0);
/* set tx fifo information */
bmwrite(dev, TXTH, 4); /* 4 octets before tx starts */
bmwrite(dev, TXFIFOCSR, 0); /* first disable txFIFO */
bmwrite(dev, TXFIFOCSR, TxFIFOEnable );
/* set rx fifo information */
bmwrite(dev, RXFIFOCSR, 0); /* first disable rxFIFO */
bmwrite(dev, RXFIFOCSR, RxFIFOEnable );
//bmwrite(dev, TXCFG, TxMACEnable); /* TxNeverGiveUp maybe later */
bmread(dev, STATUS); /* read it just to clear it */
/* zero out the chip Hash Filter registers */
for (i=0; i<4; i++) bp->hash_table_mask[i] = 0;
bmwrite(dev, BHASH3, bp->hash_table_mask[0]); /* bits 15 - 0 */
bmwrite(dev, BHASH2, bp->hash_table_mask[1]); /* bits 31 - 16 */
bmwrite(dev, BHASH1, bp->hash_table_mask[2]); /* bits 47 - 32 */
bmwrite(dev, BHASH0, bp->hash_table_mask[3]); /* bits 63 - 48 */
pWord16 = (unsigned short *)dev->dev_addr;
bmwrite(dev, MADD0, *pWord16++);
bmwrite(dev, MADD1, *pWord16++);
bmwrite(dev, MADD2, *pWord16);
bmwrite(dev, RXCFG, RxCRCNoStrip | RxHashFilterEnable | RxRejectOwnPackets);
bmwrite(dev, INTDISABLE, EnableNormal);
}
#if 0
static void
bmac_disable_interrupts(struct net_device *dev)
{
bmwrite(dev, INTDISABLE, DisableAll);
}
static void
bmac_enable_interrupts(struct net_device *dev)
{
bmwrite(dev, INTDISABLE, EnableNormal);
}
#endif
static void
bmac_start_chip(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
unsigned short oldConfig;
/* enable rx dma channel */
dbdma_continue(rd);
oldConfig = bmread(dev, TXCFG);
bmwrite(dev, TXCFG, oldConfig | TxMACEnable );
/* turn on rx plus any other bits already on (promiscuous possibly) */
oldConfig = bmread(dev, RXCFG);
bmwrite(dev, RXCFG, oldConfig | RxMACEnable );
udelay(20000);
}
static void
bmac_init_phy(struct net_device *dev)
{
unsigned int addr;
struct bmac_data *bp = netdev_priv(dev);
printk(KERN_DEBUG "phy registers:");
for (addr = 0; addr < 32; ++addr) {
if ((addr & 7) == 0)
printk(KERN_DEBUG);
printk(KERN_CONT " %.4x", bmac_mif_read(dev, addr));
}
printk(KERN_CONT "\n");
if (bp->is_bmac_plus) {
unsigned int capable, ctrl;
ctrl = bmac_mif_read(dev, 0);
capable = ((bmac_mif_read(dev, 1) & 0xf800) >> 6) | 1;
if (bmac_mif_read(dev, 4) != capable ||
(ctrl & 0x1000) == 0) {
bmac_mif_write(dev, 4, capable);
bmac_mif_write(dev, 0, 0x1200);
} else
bmac_mif_write(dev, 0, 0x1000);
}
}
static void bmac_init_chip(struct net_device *dev)
{
bmac_init_phy(dev);
bmac_init_registers(dev);
}
#ifdef CONFIG_PM
static int bmac_suspend(struct macio_dev *mdev, pm_message_t state)
{
struct net_device* dev = macio_get_drvdata(mdev);
struct bmac_data *bp = netdev_priv(dev);
unsigned long flags;
unsigned short config;
int i;
netif_device_detach(dev);
/* prolly should wait for dma to finish & turn off the chip */
spin_lock_irqsave(&bp->lock, flags);
if (bp->timeout_active) {
del_timer(&bp->tx_timeout);
bp->timeout_active = 0;
}
disable_irq(dev->irq);
disable_irq(bp->tx_dma_intr);
disable_irq(bp->rx_dma_intr);
bp->sleeping = 1;
spin_unlock_irqrestore(&bp->lock, flags);
if (bp->opened) {
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
volatile struct dbdma_regs __iomem *td = bp->tx_dma;
config = bmread(dev, RXCFG);
bmwrite(dev, RXCFG, (config & ~RxMACEnable));
config = bmread(dev, TXCFG);
bmwrite(dev, TXCFG, (config & ~TxMACEnable));
bmwrite(dev, INTDISABLE, DisableAll); /* disable all intrs */
/* disable rx and tx dma */
st_le32(&rd->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE)); /* clear run bit */
st_le32(&td->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE)); /* clear run bit */
/* free some skb's */
for (i=0; i<N_RX_RING; i++) {
if (bp->rx_bufs[i] != NULL) {
dev_kfree_skb(bp->rx_bufs[i]);
bp->rx_bufs[i] = NULL;
}
}
for (i = 0; i<N_TX_RING; i++) {
if (bp->tx_bufs[i] != NULL) {
dev_kfree_skb(bp->tx_bufs[i]);
bp->tx_bufs[i] = NULL;
}
}
}
pmac_call_feature(PMAC_FTR_BMAC_ENABLE, macio_get_of_node(bp->mdev), 0, 0);
return 0;
}
static int bmac_resume(struct macio_dev *mdev)
{
struct net_device* dev = macio_get_drvdata(mdev);
struct bmac_data *bp = netdev_priv(dev);
/* see if this is enough */
if (bp->opened)
bmac_reset_and_enable(dev);
enable_irq(dev->irq);
enable_irq(bp->tx_dma_intr);
enable_irq(bp->rx_dma_intr);
netif_device_attach(dev);
return 0;
}
#endif /* CONFIG_PM */
static int bmac_set_address(struct net_device *dev, void *addr)
{
struct bmac_data *bp = netdev_priv(dev);
unsigned char *p = addr;
unsigned short *pWord16;
unsigned long flags;
int i;
XXDEBUG(("bmac: enter set_address\n"));
spin_lock_irqsave(&bp->lock, flags);
for (i = 0; i < 6; ++i) {
dev->dev_addr[i] = p[i];
}
/* load up the hardware address */
pWord16 = (unsigned short *)dev->dev_addr;
bmwrite(dev, MADD0, *pWord16++);
bmwrite(dev, MADD1, *pWord16++);
bmwrite(dev, MADD2, *pWord16);
spin_unlock_irqrestore(&bp->lock, flags);
XXDEBUG(("bmac: exit set_address\n"));
return 0;
}
static inline void bmac_set_timeout(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
unsigned long flags;
spin_lock_irqsave(&bp->lock, flags);
if (bp->timeout_active)
del_timer(&bp->tx_timeout);
bp->tx_timeout.expires = jiffies + TX_TIMEOUT;
bp->tx_timeout.function = bmac_tx_timeout;
bp->tx_timeout.data = (unsigned long) dev;
add_timer(&bp->tx_timeout);
bp->timeout_active = 1;
spin_unlock_irqrestore(&bp->lock, flags);
}
static void
bmac_construct_xmt(struct sk_buff *skb, volatile struct dbdma_cmd *cp)
{
void *vaddr;
unsigned long baddr;
unsigned long len;
len = skb->len;
vaddr = skb->data;
baddr = virt_to_bus(vaddr);
dbdma_setcmd(cp, (OUTPUT_LAST | INTR_ALWAYS | WAIT_IFCLR), len, baddr, 0);
}
static void
bmac_construct_rxbuff(struct sk_buff *skb, volatile struct dbdma_cmd *cp)
{
unsigned char *addr = skb? skb->data: bmac_emergency_rxbuf;
dbdma_setcmd(cp, (INPUT_LAST | INTR_ALWAYS), RX_BUFLEN,
virt_to_bus(addr), 0);
}
static void
bmac_init_tx_ring(struct bmac_data *bp)
{
volatile struct dbdma_regs __iomem *td = bp->tx_dma;
memset((char *)bp->tx_cmds, 0, (N_TX_RING+1) * sizeof(struct dbdma_cmd));
bp->tx_empty = 0;
bp->tx_fill = 0;
bp->tx_fullup = 0;
/* put a branch at the end of the tx command list */
dbdma_setcmd(&bp->tx_cmds[N_TX_RING],
(DBDMA_NOP | BR_ALWAYS), 0, 0, virt_to_bus(bp->tx_cmds));
/* reset tx dma */
dbdma_reset(td);
out_le32(&td->wait_sel, 0x00200020);
out_le32(&td->cmdptr, virt_to_bus(bp->tx_cmds));
}
static int
bmac_init_rx_ring(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
int i;
struct sk_buff *skb;
/* initialize list of sk_buffs for receiving and set up recv dma */
memset((char *)bp->rx_cmds, 0,
(N_RX_RING + 1) * sizeof(struct dbdma_cmd));
for (i = 0; i < N_RX_RING; i++) {
if ((skb = bp->rx_bufs[i]) == NULL) {
bp->rx_bufs[i] = skb = netdev_alloc_skb(dev, RX_BUFLEN + 2);
if (skb != NULL)
skb_reserve(skb, 2);
}
bmac_construct_rxbuff(skb, &bp->rx_cmds[i]);
}
bp->rx_empty = 0;
bp->rx_fill = i;
/* Put a branch back to the beginning of the receive command list */
dbdma_setcmd(&bp->rx_cmds[N_RX_RING],
(DBDMA_NOP | BR_ALWAYS), 0, 0, virt_to_bus(bp->rx_cmds));
/* start rx dma */
dbdma_reset(rd);
out_le32(&rd->cmdptr, virt_to_bus(bp->rx_cmds));
return 1;
}
static int bmac_transmit_packet(struct sk_buff *skb, struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *td = bp->tx_dma;
int i;
/* see if there's a free slot in the tx ring */
/* XXDEBUG(("bmac_xmit_start: empty=%d fill=%d\n", */
/* bp->tx_empty, bp->tx_fill)); */
i = bp->tx_fill + 1;
if (i >= N_TX_RING)
i = 0;
if (i == bp->tx_empty) {
netif_stop_queue(dev);
bp->tx_fullup = 1;
XXDEBUG(("bmac_transmit_packet: tx ring full\n"));
return -1; /* can't take it at the moment */
}
dbdma_setcmd(&bp->tx_cmds[i], DBDMA_STOP, 0, 0, 0);
bmac_construct_xmt(skb, &bp->tx_cmds[bp->tx_fill]);
bp->tx_bufs[bp->tx_fill] = skb;
bp->tx_fill = i;
dev->stats.tx_bytes += skb->len;
dbdma_continue(td);
return 0;
}
static int rxintcount;
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 bmac_rxdma_intr(int irq, void *dev_id)
{
struct net_device *dev = (struct net_device *) dev_id;
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
volatile struct dbdma_cmd *cp;
int i, nb, stat;
struct sk_buff *skb;
unsigned int residual;
int last;
unsigned long flags;
spin_lock_irqsave(&bp->lock, flags);
if (++rxintcount < 10) {
XXDEBUG(("bmac_rxdma_intr\n"));
}
last = -1;
i = bp->rx_empty;
while (1) {
cp = &bp->rx_cmds[i];
stat = ld_le16(&cp->xfer_status);
residual = ld_le16(&cp->res_count);
if ((stat & ACTIVE) == 0)
break;
nb = RX_BUFLEN - residual - 2;
if (nb < (ETHERMINPACKET - ETHERCRC)) {
skb = NULL;
dev->stats.rx_length_errors++;
dev->stats.rx_errors++;
} else {
skb = bp->rx_bufs[i];
bp->rx_bufs[i] = NULL;
}
if (skb != NULL) {
nb -= ETHERCRC;
skb_put(skb, nb);
skb->protocol = eth_type_trans(skb, dev);
netif_rx(skb);
++dev->stats.rx_packets;
dev->stats.rx_bytes += nb;
} else {
++dev->stats.rx_dropped;
}
if ((skb = bp->rx_bufs[i]) == NULL) {
bp->rx_bufs[i] = skb = netdev_alloc_skb(dev, RX_BUFLEN + 2);
if (skb != NULL)
skb_reserve(bp->rx_bufs[i], 2);
}
bmac_construct_rxbuff(skb, &bp->rx_cmds[i]);
st_le16(&cp->res_count, 0);
st_le16(&cp->xfer_status, 0);
last = i;
if (++i >= N_RX_RING) i = 0;
}
if (last != -1) {
bp->rx_fill = last;
bp->rx_empty = i;
}
dbdma_continue(rd);
spin_unlock_irqrestore(&bp->lock, flags);
if (rxintcount < 10) {
XXDEBUG(("bmac_rxdma_intr done\n"));
}
return IRQ_HANDLED;
}
static int txintcount;
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 bmac_txdma_intr(int irq, void *dev_id)
{
struct net_device *dev = (struct net_device *) dev_id;
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_cmd *cp;
int stat;
unsigned long flags;
spin_lock_irqsave(&bp->lock, flags);
if (txintcount++ < 10) {
XXDEBUG(("bmac_txdma_intr\n"));
}
/* del_timer(&bp->tx_timeout); */
/* bp->timeout_active = 0; */
while (1) {
cp = &bp->tx_cmds[bp->tx_empty];
stat = ld_le16(&cp->xfer_status);
if (txintcount < 10) {
XXDEBUG(("bmac_txdma_xfer_stat=%#0x\n", stat));
}
if (!(stat & ACTIVE)) {
/*
* status field might not have been filled by DBDMA
*/
if (cp == bus_to_virt(in_le32(&bp->tx_dma->cmdptr)))
break;
}
if (bp->tx_bufs[bp->tx_empty]) {
++dev->stats.tx_packets;
dev_kfree_skb_irq(bp->tx_bufs[bp->tx_empty]);
}
bp->tx_bufs[bp->tx_empty] = NULL;
bp->tx_fullup = 0;
netif_wake_queue(dev);
if (++bp->tx_empty >= N_TX_RING)
bp->tx_empty = 0;
if (bp->tx_empty == bp->tx_fill)
break;
}
spin_unlock_irqrestore(&bp->lock, flags);
if (txintcount < 10) {
XXDEBUG(("bmac_txdma_intr done->bmac_start\n"));
}
bmac_start(dev);
return IRQ_HANDLED;
}
#ifndef SUNHME_MULTICAST
/* Real fast bit-reversal algorithm, 6-bit values */
static int reverse6[64] = {
0x0,0x20,0x10,0x30,0x8,0x28,0x18,0x38,
0x4,0x24,0x14,0x34,0xc,0x2c,0x1c,0x3c,
0x2,0x22,0x12,0x32,0xa,0x2a,0x1a,0x3a,
0x6,0x26,0x16,0x36,0xe,0x2e,0x1e,0x3e,
0x1,0x21,0x11,0x31,0x9,0x29,0x19,0x39,
0x5,0x25,0x15,0x35,0xd,0x2d,0x1d,0x3d,
0x3,0x23,0x13,0x33,0xb,0x2b,0x1b,0x3b,
0x7,0x27,0x17,0x37,0xf,0x2f,0x1f,0x3f
};
static unsigned int
crc416(unsigned int curval, unsigned short nxtval)
{
register unsigned int counter, cur = curval, next = nxtval;
register int high_crc_set, low_data_set;
/* Swap bytes */
next = ((next & 0x00FF) << 8) | (next >> 8);
/* Compute bit-by-bit */
for (counter = 0; counter < 16; ++counter) {
/* is high CRC bit set? */
if ((cur & 0x80000000) == 0) high_crc_set = 0;
else high_crc_set = 1;
cur = cur << 1;
if ((next & 0x0001) == 0) low_data_set = 0;
else low_data_set = 1;
next = next >> 1;
/* do the XOR */
if (high_crc_set ^ low_data_set) cur = cur ^ ENET_CRCPOLY;
}
return cur;
}
static unsigned int
bmac_crc(unsigned short *address)
{
unsigned int newcrc;
XXDEBUG(("bmac_crc: addr=%#04x, %#04x, %#04x\n", *address, address[1], address[2]));
newcrc = crc416(0xffffffff, *address); /* address bits 47 - 32 */
newcrc = crc416(newcrc, address[1]); /* address bits 31 - 16 */
newcrc = crc416(newcrc, address[2]); /* address bits 15 - 0 */
return(newcrc);
}
/*
* Add requested mcast addr to BMac's hash table filter.
*
*/
static void
bmac_addhash(struct bmac_data *bp, unsigned char *addr)
{
unsigned int crc;
unsigned short mask;
if (!(*addr)) return;
crc = bmac_crc((unsigned short *)addr) & 0x3f; /* Big-endian alert! */
crc = reverse6[crc]; /* Hyperfast bit-reversing algorithm */
if (bp->hash_use_count[crc]++) return; /* This bit is already set */
mask = crc % 16;
mask = (unsigned char)1 << mask;
bp->hash_use_count[crc/16] |= mask;
}
static void
bmac_removehash(struct bmac_data *bp, unsigned char *addr)
{
unsigned int crc;
unsigned char mask;
/* Now, delete the address from the filter copy, as indicated */
crc = bmac_crc((unsigned short *)addr) & 0x3f; /* Big-endian alert! */
crc = reverse6[crc]; /* Hyperfast bit-reversing algorithm */
if (bp->hash_use_count[crc] == 0) return; /* That bit wasn't in use! */
if (--bp->hash_use_count[crc]) return; /* That bit is still in use */
mask = crc % 16;
mask = ((unsigned char)1 << mask) ^ 0xffff; /* To turn off bit */
bp->hash_table_mask[crc/16] &= mask;
}
/*
* Sync the adapter with the software copy of the multicast mask
* (logical address filter).
*/
static void
bmac_rx_off(struct net_device *dev)
{
unsigned short rx_cfg;
rx_cfg = bmread(dev, RXCFG);
rx_cfg &= ~RxMACEnable;
bmwrite(dev, RXCFG, rx_cfg);
do {
rx_cfg = bmread(dev, RXCFG);
} while (rx_cfg & RxMACEnable);
}
unsigned short
bmac_rx_on(struct net_device *dev, int hash_enable, int promisc_enable)
{
unsigned short rx_cfg;
rx_cfg = bmread(dev, RXCFG);
rx_cfg |= RxMACEnable;
if (hash_enable) rx_cfg |= RxHashFilterEnable;
else rx_cfg &= ~RxHashFilterEnable;
if (promisc_enable) rx_cfg |= RxPromiscEnable;
else rx_cfg &= ~RxPromiscEnable;
bmwrite(dev, RXRST, RxResetValue);
bmwrite(dev, RXFIFOCSR, 0); /* first disable rxFIFO */
bmwrite(dev, RXFIFOCSR, RxFIFOEnable );
bmwrite(dev, RXCFG, rx_cfg );
return rx_cfg;
}
static void
bmac_update_hash_table_mask(struct net_device *dev, struct bmac_data *bp)
{
bmwrite(dev, BHASH3, bp->hash_table_mask[0]); /* bits 15 - 0 */
bmwrite(dev, BHASH2, bp->hash_table_mask[1]); /* bits 31 - 16 */
bmwrite(dev, BHASH1, bp->hash_table_mask[2]); /* bits 47 - 32 */
bmwrite(dev, BHASH0, bp->hash_table_mask[3]); /* bits 63 - 48 */
}
#if 0
static void
bmac_add_multi(struct net_device *dev,
struct bmac_data *bp, unsigned char *addr)
{
/* XXDEBUG(("bmac: enter bmac_add_multi\n")); */
bmac_addhash(bp, addr);
bmac_rx_off(dev);
bmac_update_hash_table_mask(dev, bp);
bmac_rx_on(dev, 1, (dev->flags & IFF_PROMISC)? 1 : 0);
/* XXDEBUG(("bmac: exit bmac_add_multi\n")); */
}
static void
bmac_remove_multi(struct net_device *dev,
struct bmac_data *bp, unsigned char *addr)
{
bmac_removehash(bp, addr);
bmac_rx_off(dev);
bmac_update_hash_table_mask(dev, bp);
bmac_rx_on(dev, 1, (dev->flags & IFF_PROMISC)? 1 : 0);
}
#endif
/* 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.
*/
static void bmac_set_multicast(struct net_device *dev)
{
struct netdev_hw_addr *ha;
struct bmac_data *bp = netdev_priv(dev);
int num_addrs = netdev_mc_count(dev);
unsigned short rx_cfg;
int i;
if (bp->sleeping)
return;
XXDEBUG(("bmac: enter bmac_set_multicast, n_addrs=%d\n", num_addrs));
if((dev->flags & IFF_ALLMULTI) || (netdev_mc_count(dev) > 64)) {
for (i=0; i<4; i++) bp->hash_table_mask[i] = 0xffff;
bmac_update_hash_table_mask(dev, bp);
rx_cfg = bmac_rx_on(dev, 1, 0);
XXDEBUG(("bmac: all multi, rx_cfg=%#08x\n"));
} else if ((dev->flags & IFF_PROMISC) || (num_addrs < 0)) {
rx_cfg = bmread(dev, RXCFG);
rx_cfg |= RxPromiscEnable;
bmwrite(dev, RXCFG, rx_cfg);
rx_cfg = bmac_rx_on(dev, 0, 1);
XXDEBUG(("bmac: promisc mode enabled, rx_cfg=%#08x\n", rx_cfg));
} else {
for (i=0; i<4; i++) bp->hash_table_mask[i] = 0;
for (i=0; i<64; i++) bp->hash_use_count[i] = 0;
if (num_addrs == 0) {
rx_cfg = bmac_rx_on(dev, 0, 0);
XXDEBUG(("bmac: multi disabled, rx_cfg=%#08x\n", rx_cfg));
} else {
netdev_for_each_mc_addr(ha, dev)
bmac_addhash(bp, ha->addr);
bmac_update_hash_table_mask(dev, bp);
rx_cfg = bmac_rx_on(dev, 1, 0);
XXDEBUG(("bmac: multi enabled, rx_cfg=%#08x\n", rx_cfg));
}
}
/* XXDEBUG(("bmac: exit bmac_set_multicast\n")); */
}
#else /* ifdef SUNHME_MULTICAST */
/* The version of set_multicast below was lifted from sunhme.c */
static void bmac_set_multicast(struct net_device *dev)
{
struct netdev_hw_addr *ha;
int i;
unsigned short rx_cfg;
u32 crc;
if((dev->flags & IFF_ALLMULTI) || (netdev_mc_count(dev) > 64)) {
bmwrite(dev, BHASH0, 0xffff);
bmwrite(dev, BHASH1, 0xffff);
bmwrite(dev, BHASH2, 0xffff);
bmwrite(dev, BHASH3, 0xffff);
} else if(dev->flags & IFF_PROMISC) {
rx_cfg = bmread(dev, RXCFG);
rx_cfg |= RxPromiscEnable;
bmwrite(dev, RXCFG, rx_cfg);
} else {
u16 hash_table[4];
rx_cfg = bmread(dev, RXCFG);
rx_cfg &= ~RxPromiscEnable;
bmwrite(dev, RXCFG, rx_cfg);
for(i = 0; i < 4; i++) hash_table[i] = 0;
netdev_for_each_mc_addr(ha, dev) {
crc = ether_crc_le(6, ha->addr);
crc >>= 26;
hash_table[crc >> 4] |= 1 << (crc & 0xf);
}
bmwrite(dev, BHASH0, hash_table[0]);
bmwrite(dev, BHASH1, hash_table[1]);
bmwrite(dev, BHASH2, hash_table[2]);
bmwrite(dev, BHASH3, hash_table[3]);
}
}
#endif /* SUNHME_MULTICAST */
static int miscintcount;
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 bmac_misc_intr(int irq, void *dev_id)
{
struct net_device *dev = (struct net_device *) dev_id;
unsigned int status = bmread(dev, STATUS);
if (miscintcount++ < 10) {
XXDEBUG(("bmac_misc_intr\n"));
}
/* XXDEBUG(("bmac_misc_intr, status=%#08x\n", status)); */
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
/* bmac_txdma_intr_inner(irq, dev_id); */
/* if (status & FrameReceived) dev->stats.rx_dropped++; */
if (status & RxErrorMask) dev->stats.rx_errors++;
if (status & RxCRCCntExp) dev->stats.rx_crc_errors++;
if (status & RxLenCntExp) dev->stats.rx_length_errors++;
if (status & RxOverFlow) dev->stats.rx_over_errors++;
if (status & RxAlignCntExp) dev->stats.rx_frame_errors++;
/* if (status & FrameSent) dev->stats.tx_dropped++; */
if (status & TxErrorMask) dev->stats.tx_errors++;
if (status & TxUnderrun) dev->stats.tx_fifo_errors++;
if (status & TxNormalCollExp) dev->stats.collisions++;
return IRQ_HANDLED;
}
/*
* Procedure for reading EEPROM
*/
#define SROMAddressLength 5
#define DataInOn 0x0008
#define DataInOff 0x0000
#define Clk 0x0002
#define ChipSelect 0x0001
#define SDIShiftCount 3
#define SD0ShiftCount 2
#define DelayValue 1000 /* number of microseconds */
#define SROMStartOffset 10 /* this is in words */
#define SROMReadCount 3 /* number of words to read from SROM */
#define SROMAddressBits 6
#define EnetAddressOffset 20
static unsigned char
bmac_clock_out_bit(struct net_device *dev)
{
unsigned short data;
unsigned short val;
bmwrite(dev, SROMCSR, ChipSelect | Clk);
udelay(DelayValue);
data = bmread(dev, SROMCSR);
udelay(DelayValue);
val = (data >> SD0ShiftCount) & 1;
bmwrite(dev, SROMCSR, ChipSelect);
udelay(DelayValue);
return val;
}
static void
bmac_clock_in_bit(struct net_device *dev, unsigned int val)
{
unsigned short data;
if (val != 0 && val != 1) return;
data = (val << SDIShiftCount);
bmwrite(dev, SROMCSR, data | ChipSelect );
udelay(DelayValue);
bmwrite(dev, SROMCSR, data | ChipSelect | Clk );
udelay(DelayValue);
bmwrite(dev, SROMCSR, data | ChipSelect);
udelay(DelayValue);
}
static void
reset_and_select_srom(struct net_device *dev)
{
/* first reset */
bmwrite(dev, SROMCSR, 0);
udelay(DelayValue);
/* send it the read command (110) */
bmac_clock_in_bit(dev, 1);
bmac_clock_in_bit(dev, 1);
bmac_clock_in_bit(dev, 0);
}
static unsigned short
read_srom(struct net_device *dev, unsigned int addr, unsigned int addr_len)
{
unsigned short data, val;
int i;
/* send out the address we want to read from */
for (i = 0; i < addr_len; i++) {
val = addr >> (addr_len-i-1);
bmac_clock_in_bit(dev, val & 1);
}
/* Now read in the 16-bit data */
data = 0;
for (i = 0; i < 16; i++) {
val = bmac_clock_out_bit(dev);
data <<= 1;
data |= val;
}
bmwrite(dev, SROMCSR, 0);
return data;
}
/*
* It looks like Cogent and SMC use different methods for calculating
* checksums. What a pain..
*/
static int
bmac_verify_checksum(struct net_device *dev)
{
unsigned short data, storedCS;
reset_and_select_srom(dev);
data = read_srom(dev, 3, SROMAddressBits);
storedCS = ((data >> 8) & 0x0ff) | ((data << 8) & 0xff00);
return 0;
}
static void
bmac_get_station_address(struct net_device *dev, unsigned char *ea)
{
int i;
unsigned short data;
for (i = 0; i < 6; i++)
{
reset_and_select_srom(dev);
data = read_srom(dev, i + EnetAddressOffset/2, SROMAddressBits);
ea[2*i] = bitrev8(data & 0x0ff);
ea[2*i+1] = bitrev8((data >> 8) & 0x0ff);
}
}
static void bmac_reset_and_enable(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
unsigned long flags;
struct sk_buff *skb;
unsigned char *data;
spin_lock_irqsave(&bp->lock, flags);
bmac_enable_and_reset_chip(dev);
bmac_init_tx_ring(bp);
bmac_init_rx_ring(dev);
bmac_init_chip(dev);
bmac_start_chip(dev);
bmwrite(dev, INTDISABLE, EnableNormal);
bp->sleeping = 0;
/*
* It seems that the bmac can't receive until it's transmitted
* a packet. So we give it a dummy packet to transmit.
*/
skb = netdev_alloc_skb(dev, ETHERMINPACKET);
if (skb != NULL) {
data = skb_put(skb, ETHERMINPACKET);
memset(data, 0, ETHERMINPACKET);
memcpy(data, dev->dev_addr, 6);
memcpy(data+6, dev->dev_addr, 6);
bmac_transmit_packet(skb, dev);
}
spin_unlock_irqrestore(&bp->lock, flags);
}
static const struct ethtool_ops bmac_ethtool_ops = {
.get_link = ethtool_op_get_link,
};
static const struct net_device_ops bmac_netdev_ops = {
.ndo_open = bmac_open,
.ndo_stop = bmac_close,
.ndo_start_xmit = bmac_output,
.ndo_set_rx_mode = bmac_set_multicast,
.ndo_set_mac_address = bmac_set_address,
.ndo_change_mtu = eth_change_mtu,
.ndo_validate_addr = eth_validate_addr,
};
static int bmac_probe(struct macio_dev *mdev, const struct of_device_id *match)
{
int j, rev, ret;
struct bmac_data *bp;
const unsigned char *prop_addr;
unsigned char addr[6];
struct net_device *dev;
int is_bmac_plus = ((int)match->data) != 0;
if (macio_resource_count(mdev) != 3 || macio_irq_count(mdev) != 3) {
printk(KERN_ERR "BMAC: can't use, need 3 addrs and 3 intrs\n");
return -ENODEV;
}
prop_addr = of_get_property(macio_get_of_node(mdev),
"mac-address", NULL);
if (prop_addr == NULL) {
prop_addr = of_get_property(macio_get_of_node(mdev),
"local-mac-address", NULL);
if (prop_addr == NULL) {
printk(KERN_ERR "BMAC: Can't get mac-address\n");
return -ENODEV;
}
}
memcpy(addr, prop_addr, sizeof(addr));
dev = alloc_etherdev(PRIV_BYTES);
if (!dev)
return -ENOMEM;
bp = netdev_priv(dev);
SET_NETDEV_DEV(dev, &mdev->ofdev.dev);
macio_set_drvdata(mdev, dev);
bp->mdev = mdev;
spin_lock_init(&bp->lock);
if (macio_request_resources(mdev, "bmac")) {
printk(KERN_ERR "BMAC: can't request IO resource !\n");
goto out_free;
}
dev->base_addr = (unsigned long)
ioremap(macio_resource_start(mdev, 0), macio_resource_len(mdev, 0));
if (dev->base_addr == 0)
goto out_release;
dev->irq = macio_irq(mdev, 0);
bmac_enable_and_reset_chip(dev);
bmwrite(dev, INTDISABLE, DisableAll);
rev = addr[0] == 0 && addr[1] == 0xA0;
for (j = 0; j < 6; ++j)
dev->dev_addr[j] = rev ? bitrev8(addr[j]): addr[j];
/* Enable chip without interrupts for now */
bmac_enable_and_reset_chip(dev);
bmwrite(dev, INTDISABLE, DisableAll);
dev->netdev_ops = &bmac_netdev_ops;
dev->ethtool_ops = &bmac_ethtool_ops;
bmac_get_station_address(dev, addr);
if (bmac_verify_checksum(dev) != 0)
goto err_out_iounmap;
bp->is_bmac_plus = is_bmac_plus;
bp->tx_dma = ioremap(macio_resource_start(mdev, 1), macio_resource_len(mdev, 1));
if (!bp->tx_dma)
goto err_out_iounmap;
bp->tx_dma_intr = macio_irq(mdev, 1);
bp->rx_dma = ioremap(macio_resource_start(mdev, 2), macio_resource_len(mdev, 2));
if (!bp->rx_dma)
goto err_out_iounmap_tx;
bp->rx_dma_intr = macio_irq(mdev, 2);
bp->tx_cmds = (volatile struct dbdma_cmd *) DBDMA_ALIGN(bp + 1);
bp->rx_cmds = bp->tx_cmds + N_TX_RING + 1;
bp->queue = (struct sk_buff_head *)(bp->rx_cmds + N_RX_RING + 1);
skb_queue_head_init(bp->queue);
init_timer(&bp->tx_timeout);
ret = request_irq(dev->irq, bmac_misc_intr, 0, "BMAC-misc", dev);
if (ret) {
printk(KERN_ERR "BMAC: can't get irq %d\n", dev->irq);
goto err_out_iounmap_rx;
}
ret = request_irq(bp->tx_dma_intr, bmac_txdma_intr, 0, "BMAC-txdma", dev);
if (ret) {
printk(KERN_ERR "BMAC: can't get irq %d\n", bp->tx_dma_intr);
goto err_out_irq0;
}
ret = request_irq(bp->rx_dma_intr, bmac_rxdma_intr, 0, "BMAC-rxdma", dev);
if (ret) {
printk(KERN_ERR "BMAC: can't get irq %d\n", bp->rx_dma_intr);
goto err_out_irq1;
}
/* Mask chip interrupts and disable chip, will be
* re-enabled on open()
*/
disable_irq(dev->irq);
pmac_call_feature(PMAC_FTR_BMAC_ENABLE, macio_get_of_node(bp->mdev), 0, 0);
if (register_netdev(dev) != 0) {
printk(KERN_ERR "BMAC: Ethernet registration failed\n");
goto err_out_irq2;
}
printk(KERN_INFO "%s: BMAC%s at %pM",
dev->name, (is_bmac_plus ? "+" : ""), dev->dev_addr);
XXDEBUG((", base_addr=%#0lx", dev->base_addr));
printk("\n");
return 0;
err_out_irq2:
free_irq(bp->rx_dma_intr, dev);
err_out_irq1:
free_irq(bp->tx_dma_intr, dev);
err_out_irq0:
free_irq(dev->irq, dev);
err_out_iounmap_rx:
iounmap(bp->rx_dma);
err_out_iounmap_tx:
iounmap(bp->tx_dma);
err_out_iounmap:
iounmap((void __iomem *)dev->base_addr);
out_release:
macio_release_resources(mdev);
out_free:
pmac_call_feature(PMAC_FTR_BMAC_ENABLE, macio_get_of_node(bp->mdev), 0, 0);
free_netdev(dev);
return -ENODEV;
}
static int bmac_open(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
/* XXDEBUG(("bmac: enter open\n")); */
/* reset the chip */
bp->opened = 1;
bmac_reset_and_enable(dev);
enable_irq(dev->irq);
return 0;
}
static int bmac_close(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
volatile struct dbdma_regs __iomem *td = bp->tx_dma;
unsigned short config;
int i;
bp->sleeping = 1;
/* disable rx and tx */
config = bmread(dev, RXCFG);
bmwrite(dev, RXCFG, (config & ~RxMACEnable));
config = bmread(dev, TXCFG);
bmwrite(dev, TXCFG, (config & ~TxMACEnable));
bmwrite(dev, INTDISABLE, DisableAll); /* disable all intrs */
/* disable rx and tx dma */
st_le32(&rd->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE)); /* clear run bit */
st_le32(&td->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE)); /* clear run bit */
/* free some skb's */
XXDEBUG(("bmac: free rx bufs\n"));
for (i=0; i<N_RX_RING; i++) {
if (bp->rx_bufs[i] != NULL) {
dev_kfree_skb(bp->rx_bufs[i]);
bp->rx_bufs[i] = NULL;
}
}
XXDEBUG(("bmac: free tx bufs\n"));
for (i = 0; i<N_TX_RING; i++) {
if (bp->tx_bufs[i] != NULL) {
dev_kfree_skb(bp->tx_bufs[i]);
bp->tx_bufs[i] = NULL;
}
}
XXDEBUG(("bmac: all bufs freed\n"));
bp->opened = 0;
disable_irq(dev->irq);
pmac_call_feature(PMAC_FTR_BMAC_ENABLE, macio_get_of_node(bp->mdev), 0, 0);
return 0;
}
static void
bmac_start(struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
int i;
struct sk_buff *skb;
unsigned long flags;
if (bp->sleeping)
return;
spin_lock_irqsave(&bp->lock, flags);
while (1) {
i = bp->tx_fill + 1;
if (i >= N_TX_RING)
i = 0;
if (i == bp->tx_empty)
break;
skb = skb_dequeue(bp->queue);
if (skb == NULL)
break;
bmac_transmit_packet(skb, dev);
}
spin_unlock_irqrestore(&bp->lock, flags);
}
static int
bmac_output(struct sk_buff *skb, struct net_device *dev)
{
struct bmac_data *bp = netdev_priv(dev);
skb_queue_tail(bp->queue, skb);
bmac_start(dev);
return NETDEV_TX_OK;
}
static void bmac_tx_timeout(unsigned long data)
{
struct net_device *dev = (struct net_device *) data;
struct bmac_data *bp = netdev_priv(dev);
volatile struct dbdma_regs __iomem *td = bp->tx_dma;
volatile struct dbdma_regs __iomem *rd = bp->rx_dma;
volatile struct dbdma_cmd *cp;
unsigned long flags;
unsigned short config, oldConfig;
int i;
XXDEBUG(("bmac: tx_timeout called\n"));
spin_lock_irqsave(&bp->lock, flags);
bp->timeout_active = 0;
/* update various counters */
/* bmac_handle_misc_intrs(bp, 0); */
cp = &bp->tx_cmds[bp->tx_empty];
/* XXDEBUG((KERN_DEBUG "bmac: tx dmastat=%x %x runt=%d pr=%x fs=%x fc=%x\n", */
/* ld_le32(&td->status), ld_le16(&cp->xfer_status), bp->tx_bad_runt, */
/* mb->pr, mb->xmtfs, mb->fifofc)); */
/* turn off both tx and rx and reset the chip */
config = bmread(dev, RXCFG);
bmwrite(dev, RXCFG, (config & ~RxMACEnable));
config = bmread(dev, TXCFG);
bmwrite(dev, TXCFG, (config & ~TxMACEnable));
out_le32(&td->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE|ACTIVE|DEAD));
printk(KERN_ERR "bmac: transmit timeout - resetting\n");
bmac_enable_and_reset_chip(dev);
/* restart rx dma */
cp = bus_to_virt(ld_le32(&rd->cmdptr));
out_le32(&rd->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE|ACTIVE|DEAD));
out_le16(&cp->xfer_status, 0);
out_le32(&rd->cmdptr, virt_to_bus(cp));
out_le32(&rd->control, DBDMA_SET(RUN|WAKE));
/* fix up the transmit side */
XXDEBUG((KERN_DEBUG "bmac: tx empty=%d fill=%d fullup=%d\n",
bp->tx_empty, bp->tx_fill, bp->tx_fullup));
i = bp->tx_empty;
++dev->stats.tx_errors;
if (i != bp->tx_fill) {
dev_kfree_skb(bp->tx_bufs[i]);
bp->tx_bufs[i] = NULL;
if (++i >= N_TX_RING) i = 0;
bp->tx_empty = i;
}
bp->tx_fullup = 0;
netif_wake_queue(dev);
if (i != bp->tx_fill) {
cp = &bp->tx_cmds[i];
out_le16(&cp->xfer_status, 0);
out_le16(&cp->command, OUTPUT_LAST);
out_le32(&td->cmdptr, virt_to_bus(cp));
out_le32(&td->control, DBDMA_SET(RUN));
/* bmac_set_timeout(dev); */
XXDEBUG((KERN_DEBUG "bmac: starting %d\n", i));
}
/* turn it back on */
oldConfig = bmread(dev, RXCFG);
bmwrite(dev, RXCFG, oldConfig | RxMACEnable );
oldConfig = bmread(dev, TXCFG);
bmwrite(dev, TXCFG, oldConfig | TxMACEnable );
spin_unlock_irqrestore(&bp->lock, flags);
}
#if 0
static void dump_dbdma(volatile struct dbdma_cmd *cp,int count)
{
int i,*ip;
for (i=0;i< count;i++) {
ip = (int*)(cp+i);
printk("dbdma req 0x%x addr 0x%x baddr 0x%x xfer/res 0x%x\n",
ld_le32(ip+0),
ld_le32(ip+1),
ld_le32(ip+2),
ld_le32(ip+3));
}
}
#endif
#if 0
static int
bmac_proc_info(char *buffer, char **start, off_t offset, int length)
{
int len = 0;
off_t pos = 0;
off_t begin = 0;
int i;
if (bmac_devs == NULL)
return -ENOSYS;
len += sprintf(buffer, "BMAC counters & registers\n");
for (i = 0; i<N_REG_ENTRIES; i++) {
len += sprintf(buffer + len, "%s: %#08x\n",
reg_entries[i].name,
bmread(bmac_devs, reg_entries[i].reg_offset));
pos = begin + len;
if (pos < offset) {
len = 0;
begin = pos;
}
if (pos > offset+length) break;
}
*start = buffer + (offset - begin);
len -= (offset - begin);
if (len > length) len = length;
return len;
}
#endif
static int bmac_remove(struct macio_dev *mdev)
{
struct net_device *dev = macio_get_drvdata(mdev);
struct bmac_data *bp = netdev_priv(dev);
unregister_netdev(dev);
free_irq(dev->irq, dev);
free_irq(bp->tx_dma_intr, dev);
free_irq(bp->rx_dma_intr, dev);
iounmap((void __iomem *)dev->base_addr);
iounmap(bp->tx_dma);
iounmap(bp->rx_dma);
macio_release_resources(mdev);
free_netdev(dev);
return 0;
}
static struct of_device_id bmac_match[] =
{
{
.name = "bmac",
.data = (void *)0,
},
{
.type = "network",
.compatible = "bmac+",
.data = (void *)1,
},
{},
};
MODULE_DEVICE_TABLE (of, bmac_match);
static struct macio_driver bmac_driver =
{
.driver = {
.name = "bmac",
.owner = THIS_MODULE,
.of_match_table = bmac_match,
},
.probe = bmac_probe,
.remove = bmac_remove,
#ifdef CONFIG_PM
.suspend = bmac_suspend,
.resume = bmac_resume,
#endif
};
static int __init bmac_init(void)
{
if (bmac_emergency_rxbuf == NULL) {
bmac_emergency_rxbuf = kmalloc(RX_BUFLEN, GFP_KERNEL);
if (bmac_emergency_rxbuf == NULL)
return -ENOMEM;
}
return macio_register_driver(&bmac_driver);
}
static void __exit bmac_exit(void)
{
macio_unregister_driver(&bmac_driver);
kfree(bmac_emergency_rxbuf);
bmac_emergency_rxbuf = NULL;
}
MODULE_AUTHOR("Randy Gobbel/Paul Mackerras");
MODULE_DESCRIPTION("PowerMac BMAC ethernet driver.");
MODULE_LICENSE("GPL");
module_init(bmac_init);
module_exit(bmac_exit);