linux_dsm_epyc7002/drivers/net/3c527.c
Linus Torvalds 1da177e4c3 Linux-2.6.12-rc2
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!
2005-04-16 15:20:36 -07:00

1676 lines
43 KiB
C

/* 3c527.c: 3Com Etherlink/MC32 driver for Linux 2.4 and 2.6.
*
* (c) Copyright 1998 Red Hat Software Inc
* Written by Alan Cox.
* Further debugging by Carl Drougge.
* Initial SMP support by Felipe W Damasio <felipewd@terra.com.br>
* Heavily modified by Richard Procter <rnp@paradise.net.nz>
*
* Based on skeleton.c written 1993-94 by Donald Becker and ne2.c
* (for the MCA stuff) written by Wim Dumon.
*
* Thanks to 3Com for making this possible by providing me with the
* documentation.
*
* This software may be used and distributed according to the terms
* of the GNU General Public License, incorporated herein by reference.
*
*/
#define DRV_NAME "3c527"
#define DRV_VERSION "0.7-SMP"
#define DRV_RELDATE "2003/09/21"
static const char *version =
DRV_NAME ".c:v" DRV_VERSION " " DRV_RELDATE " Richard Procter <rnp@paradise.net.nz>\n";
/**
* DOC: Traps for the unwary
*
* The diagram (Figure 1-1) and the POS summary disagree with the
* "Interrupt Level" section in the manual.
*
* The manual contradicts itself when describing the minimum number
* buffers in the 'configure lists' command.
* My card accepts a buffer config of 4/4.
*
* Setting the SAV BP bit does not save bad packets, but
* only enables RX on-card stats collection.
*
* The documentation in places seems to miss things. In actual fact
* I've always eventually found everything is documented, it just
* requires careful study.
*
* DOC: Theory Of Operation
*
* The 3com 3c527 is a 32bit MCA bus mastering adapter with a large
* amount of on board intelligence that housekeeps a somewhat dumber
* Intel NIC. For performance we want to keep the transmit queue deep
* as the card can transmit packets while fetching others from main
* memory by bus master DMA. Transmission and reception are driven by
* circular buffer queues.
*
* The mailboxes can be used for controlling how the card traverses
* its buffer rings, but are used only for inital setup in this
* implementation. The exec mailbox allows a variety of commands to
* be executed. Each command must complete before the next is
* executed. Primarily we use the exec mailbox for controlling the
* multicast lists. We have to do a certain amount of interesting
* hoop jumping as the multicast list changes can occur in interrupt
* state when the card has an exec command pending. We defer such
* events until the command completion interrupt.
*
* A copy break scheme (taken from 3c59x.c) is employed whereby
* received frames exceeding a configurable length are passed
* directly to the higher networking layers without incuring a copy,
* in what amounts to a time/space trade-off.
*
* The card also keeps a large amount of statistical information
* on-board. In a perfect world, these could be used safely at no
* cost. However, lacking information to the contrary, processing
* them without races would involve so much extra complexity as to
* make it unworthwhile to do so. In the end, a hybrid SW/HW
* implementation was made necessary --- see mc32_update_stats().
*
* DOC: Notes
*
* It should be possible to use two or more cards, but at this stage
* only by loading two copies of the same module.
*
* The on-board 82586 NIC has trouble receiving multiple
* back-to-back frames and so is likely to drop packets from fast
* senders.
**/
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/if_ether.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/types.h>
#include <linux/fcntl.h>
#include <linux/interrupt.h>
#include <linux/mca-legacy.h>
#include <linux/ioport.h>
#include <linux/in.h>
#include <linux/skbuff.h>
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/wait.h>
#include <linux/ethtool.h>
#include <linux/completion.h>
#include <linux/bitops.h>
#include <asm/semaphore.h>
#include <asm/uaccess.h>
#include <asm/system.h>
#include <asm/io.h>
#include <asm/dma.h>
#include "3c527.h"
MODULE_LICENSE("GPL");
/*
* The name of the card. Is used for messages and in the requests for
* io regions, irqs and dma channels
*/
static const char* cardname = DRV_NAME;
/* use 0 for production, 1 for verification, >2 for debug */
#ifndef NET_DEBUG
#define NET_DEBUG 2
#endif
#undef DEBUG_IRQ
static unsigned int mc32_debug = NET_DEBUG;
/* The number of low I/O ports used by the ethercard. */
#define MC32_IO_EXTENT 8
/* As implemented, values must be a power-of-2 -- 4/8/16/32 */
#define TX_RING_LEN 32 /* Typically the card supports 37 */
#define RX_RING_LEN 8 /* " " " */
/* Copy break point, see above for details.
* Setting to > 1512 effectively disables this feature. */
#define RX_COPYBREAK 200 /* Value from 3c59x.c */
/* Issue the 82586 workaround command - this is for "busy lans", but
* basically means for all lans now days - has a performance (latency)
* cost, but best set. */
static const int WORKAROUND_82586=1;
/* Pointers to buffers and their on-card records */
struct mc32_ring_desc
{
volatile struct skb_header *p;
struct sk_buff *skb;
};
/* Information that needs to be kept for each board. */
struct mc32_local
{
int slot;
u32 base;
struct net_device_stats net_stats;
volatile struct mc32_mailbox *rx_box;
volatile struct mc32_mailbox *tx_box;
volatile struct mc32_mailbox *exec_box;
volatile struct mc32_stats *stats; /* Start of on-card statistics */
u16 tx_chain; /* Transmit list start offset */
u16 rx_chain; /* Receive list start offset */
u16 tx_len; /* Transmit list count */
u16 rx_len; /* Receive list count */
u16 xceiver_desired_state; /* HALTED or RUNNING */
u16 cmd_nonblocking; /* Thread is uninterested in command result */
u16 mc_reload_wait; /* A multicast load request is pending */
u32 mc_list_valid; /* True when the mclist is set */
struct mc32_ring_desc tx_ring[TX_RING_LEN]; /* Host Transmit ring */
struct mc32_ring_desc rx_ring[RX_RING_LEN]; /* Host Receive ring */
atomic_t tx_count; /* buffers left */
atomic_t tx_ring_head; /* index to tx en-queue end */
u16 tx_ring_tail; /* index to tx de-queue end */
u16 rx_ring_tail; /* index to rx de-queue end */
struct semaphore cmd_mutex; /* Serialises issuing of execute commands */
struct completion execution_cmd; /* Card has completed an execute command */
struct completion xceiver_cmd; /* Card has completed a tx or rx command */
};
/* The station (ethernet) address prefix, used for a sanity check. */
#define SA_ADDR0 0x02
#define SA_ADDR1 0x60
#define SA_ADDR2 0xAC
struct mca_adapters_t {
unsigned int id;
char *name;
};
static const struct mca_adapters_t mc32_adapters[] = {
{ 0x0041, "3COM EtherLink MC/32" },
{ 0x8EF5, "IBM High Performance Lan Adapter" },
{ 0x0000, NULL }
};
/* Macros for ring index manipulations */
static inline u16 next_rx(u16 rx) { return (rx+1)&(RX_RING_LEN-1); };
static inline u16 prev_rx(u16 rx) { return (rx-1)&(RX_RING_LEN-1); };
static inline u16 next_tx(u16 tx) { return (tx+1)&(TX_RING_LEN-1); };
/* Index to functions, as function prototypes. */
static int mc32_probe1(struct net_device *dev, int ioaddr);
static int mc32_command(struct net_device *dev, u16 cmd, void *data, int len);
static int mc32_open(struct net_device *dev);
static void mc32_timeout(struct net_device *dev);
static int mc32_send_packet(struct sk_buff *skb, struct net_device *dev);
static irqreturn_t mc32_interrupt(int irq, void *dev_id, struct pt_regs *regs);
static int mc32_close(struct net_device *dev);
static struct net_device_stats *mc32_get_stats(struct net_device *dev);
static void mc32_set_multicast_list(struct net_device *dev);
static void mc32_reset_multicast_list(struct net_device *dev);
static struct ethtool_ops netdev_ethtool_ops;
static void cleanup_card(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
unsigned slot = lp->slot;
mca_mark_as_unused(slot);
mca_set_adapter_name(slot, NULL);
free_irq(dev->irq, dev);
release_region(dev->base_addr, MC32_IO_EXTENT);
}
/**
* mc32_probe - Search for supported boards
* @unit: interface number to use
*
* Because MCA bus is a real bus and we can scan for cards we could do a
* single scan for all boards here. Right now we use the passed in device
* structure and scan for only one board. This needs fixing for modules
* in particular.
*/
struct net_device *__init mc32_probe(int unit)
{
struct net_device *dev = alloc_etherdev(sizeof(struct mc32_local));
static int current_mca_slot = -1;
int i;
int err;
if (!dev)
return ERR_PTR(-ENOMEM);
if (unit >= 0)
sprintf(dev->name, "eth%d", unit);
SET_MODULE_OWNER(dev);
/* Do not check any supplied i/o locations.
POS registers usually don't fail :) */
/* MCA cards have POS registers.
Autodetecting MCA cards is extremely simple.
Just search for the card. */
for(i = 0; (mc32_adapters[i].name != NULL); i++) {
current_mca_slot =
mca_find_unused_adapter(mc32_adapters[i].id, 0);
if(current_mca_slot != MCA_NOTFOUND) {
if(!mc32_probe1(dev, current_mca_slot))
{
mca_set_adapter_name(current_mca_slot,
mc32_adapters[i].name);
mca_mark_as_used(current_mca_slot);
err = register_netdev(dev);
if (err) {
cleanup_card(dev);
free_netdev(dev);
dev = ERR_PTR(err);
}
return dev;
}
}
}
free_netdev(dev);
return ERR_PTR(-ENODEV);
}
/**
* mc32_probe1 - Check a given slot for a board and test the card
* @dev: Device structure to fill in
* @slot: The MCA bus slot being used by this card
*
* Decode the slot data and configure the card structures. Having done this we
* can reset the card and configure it. The card does a full self test cycle
* in firmware so we have to wait for it to return and post us either a
* failure case or some addresses we use to find the board internals.
*/
static int __init mc32_probe1(struct net_device *dev, int slot)
{
static unsigned version_printed;
int i, err;
u8 POS;
u32 base;
struct mc32_local *lp = netdev_priv(dev);
static u16 mca_io_bases[]={
0x7280,0x7290,
0x7680,0x7690,
0x7A80,0x7A90,
0x7E80,0x7E90
};
static u32 mca_mem_bases[]={
0x00C0000,
0x00C4000,
0x00C8000,
0x00CC000,
0x00D0000,
0x00D4000,
0x00D8000,
0x00DC000
};
static char *failures[]={
"Processor instruction",
"Processor data bus",
"Processor data bus",
"Processor data bus",
"Adapter bus",
"ROM checksum",
"Base RAM",
"Extended RAM",
"82586 internal loopback",
"82586 initialisation failure",
"Adapter list configuration error"
};
/* Time to play MCA games */
if (mc32_debug && version_printed++ == 0)
printk(KERN_DEBUG "%s", version);
printk(KERN_INFO "%s: %s found in slot %d:", dev->name, cardname, slot);
POS = mca_read_stored_pos(slot, 2);
if(!(POS&1))
{
printk(" disabled.\n");
return -ENODEV;
}
/* Fill in the 'dev' fields. */
dev->base_addr = mca_io_bases[(POS>>1)&7];
dev->mem_start = mca_mem_bases[(POS>>4)&7];
POS = mca_read_stored_pos(slot, 4);
if(!(POS&1))
{
printk("memory window disabled.\n");
return -ENODEV;
}
POS = mca_read_stored_pos(slot, 5);
i=(POS>>4)&3;
if(i==3)
{
printk("invalid memory window.\n");
return -ENODEV;
}
i*=16384;
i+=16384;
dev->mem_end=dev->mem_start + i;
dev->irq = ((POS>>2)&3)+9;
if(!request_region(dev->base_addr, MC32_IO_EXTENT, cardname))
{
printk("io 0x%3lX, which is busy.\n", dev->base_addr);
return -EBUSY;
}
printk("io 0x%3lX irq %d mem 0x%lX (%dK)\n",
dev->base_addr, dev->irq, dev->mem_start, i/1024);
/* We ought to set the cache line size here.. */
/*
* Go PROM browsing
*/
printk("%s: Address ", dev->name);
/* Retrieve and print the ethernet address. */
for (i = 0; i < 6; i++)
{
mca_write_pos(slot, 6, i+12);
mca_write_pos(slot, 7, 0);
printk(" %2.2x", dev->dev_addr[i] = mca_read_pos(slot,3));
}
mca_write_pos(slot, 6, 0);
mca_write_pos(slot, 7, 0);
POS = mca_read_stored_pos(slot, 4);
if(POS&2)
printk(" : BNC port selected.\n");
else
printk(" : AUI port selected.\n");
POS=inb(dev->base_addr+HOST_CTRL);
POS|=HOST_CTRL_ATTN|HOST_CTRL_RESET;
POS&=~HOST_CTRL_INTE;
outb(POS, dev->base_addr+HOST_CTRL);
/* Reset adapter */
udelay(100);
/* Reset off */
POS&=~(HOST_CTRL_ATTN|HOST_CTRL_RESET);
outb(POS, dev->base_addr+HOST_CTRL);
udelay(300);
/*
* Grab the IRQ
*/
err = request_irq(dev->irq, &mc32_interrupt, SA_SHIRQ | SA_SAMPLE_RANDOM, DRV_NAME, dev);
if (err) {
release_region(dev->base_addr, MC32_IO_EXTENT);
printk(KERN_ERR "%s: unable to get IRQ %d.\n", DRV_NAME, dev->irq);
goto err_exit_ports;
}
memset(lp, 0, sizeof(struct mc32_local));
lp->slot = slot;
i=0;
base = inb(dev->base_addr);
while(base == 0xFF)
{
i++;
if(i == 1000)
{
printk(KERN_ERR "%s: failed to boot adapter.\n", dev->name);
err = -ENODEV;
goto err_exit_irq;
}
udelay(1000);
if(inb(dev->base_addr+2)&(1<<5))
base = inb(dev->base_addr);
}
if(base>0)
{
if(base < 0x0C)
printk(KERN_ERR "%s: %s%s.\n", dev->name, failures[base-1],
base<0x0A?" test failure":"");
else
printk(KERN_ERR "%s: unknown failure %d.\n", dev->name, base);
err = -ENODEV;
goto err_exit_irq;
}
base=0;
for(i=0;i<4;i++)
{
int n=0;
while(!(inb(dev->base_addr+2)&(1<<5)))
{
n++;
udelay(50);
if(n>100)
{
printk(KERN_ERR "%s: mailbox read fail (%d).\n", dev->name, i);
err = -ENODEV;
goto err_exit_irq;
}
}
base|=(inb(dev->base_addr)<<(8*i));
}
lp->exec_box=isa_bus_to_virt(dev->mem_start+base);
base=lp->exec_box->data[1]<<16|lp->exec_box->data[0];
lp->base = dev->mem_start+base;
lp->rx_box=isa_bus_to_virt(lp->base + lp->exec_box->data[2]);
lp->tx_box=isa_bus_to_virt(lp->base + lp->exec_box->data[3]);
lp->stats = isa_bus_to_virt(lp->base + lp->exec_box->data[5]);
/*
* Descriptor chains (card relative)
*/
lp->tx_chain = lp->exec_box->data[8]; /* Transmit list start offset */
lp->rx_chain = lp->exec_box->data[10]; /* Receive list start offset */
lp->tx_len = lp->exec_box->data[9]; /* Transmit list count */
lp->rx_len = lp->exec_box->data[11]; /* Receive list count */
init_MUTEX_LOCKED(&lp->cmd_mutex);
init_completion(&lp->execution_cmd);
init_completion(&lp->xceiver_cmd);
printk("%s: Firmware Rev %d. %d RX buffers, %d TX buffers. Base of 0x%08X.\n",
dev->name, lp->exec_box->data[12], lp->rx_len, lp->tx_len, lp->base);
dev->open = mc32_open;
dev->stop = mc32_close;
dev->hard_start_xmit = mc32_send_packet;
dev->get_stats = mc32_get_stats;
dev->set_multicast_list = mc32_set_multicast_list;
dev->tx_timeout = mc32_timeout;
dev->watchdog_timeo = HZ*5; /* Board does all the work */
dev->ethtool_ops = &netdev_ethtool_ops;
return 0;
err_exit_irq:
free_irq(dev->irq, dev);
err_exit_ports:
release_region(dev->base_addr, MC32_IO_EXTENT);
return err;
}
/**
* mc32_ready_poll - wait until we can feed it a command
* @dev: The device to wait for
*
* Wait until the card becomes ready to accept a command via the
* command register. This tells us nothing about the completion
* status of any pending commands and takes very little time at all.
*/
static inline void mc32_ready_poll(struct net_device *dev)
{
int ioaddr = dev->base_addr;
while(!(inb(ioaddr+HOST_STATUS)&HOST_STATUS_CRR));
}
/**
* mc32_command_nowait - send a command non blocking
* @dev: The 3c527 to issue the command to
* @cmd: The command word to write to the mailbox
* @data: A data block if the command expects one
* @len: Length of the data block
*
* Send a command from interrupt state. If there is a command
* currently being executed then we return an error of -1. It
* simply isn't viable to wait around as commands may be
* slow. This can theoretically be starved on SMP, but it's hard
* to see a realistic situation. We do not wait for the command
* to complete --- we rely on the interrupt handler to tidy up
* after us.
*/
static int mc32_command_nowait(struct net_device *dev, u16 cmd, void *data, int len)
{
struct mc32_local *lp = netdev_priv(dev);
int ioaddr = dev->base_addr;
int ret = -1;
if (down_trylock(&lp->cmd_mutex) == 0)
{
lp->cmd_nonblocking=1;
lp->exec_box->mbox=0;
lp->exec_box->mbox=cmd;
memcpy((void *)lp->exec_box->data, data, len);
barrier(); /* the memcpy forgot the volatile so be sure */
/* Send the command */
mc32_ready_poll(dev);
outb(1<<6, ioaddr+HOST_CMD);
ret = 0;
/* Interrupt handler will signal mutex on completion */
}
return ret;
}
/**
* mc32_command - send a command and sleep until completion
* @dev: The 3c527 card to issue the command to
* @cmd: The command word to write to the mailbox
* @data: A data block if the command expects one
* @len: Length of the data block
*
* Sends exec commands in a user context. This permits us to wait around
* for the replies and also to wait for the command buffer to complete
* from a previous command before we execute our command. After our
* command completes we will attempt any pending multicast reload
* we blocked off by hogging the exec buffer.
*
* You feed the card a command, you wait, it interrupts you get a
* reply. All well and good. The complication arises because you use
* commands for filter list changes which come in at bh level from things
* like IPV6 group stuff.
*/
static int mc32_command(struct net_device *dev, u16 cmd, void *data, int len)
{
struct mc32_local *lp = netdev_priv(dev);
int ioaddr = dev->base_addr;
int ret = 0;
down(&lp->cmd_mutex);
/*
* My Turn
*/
lp->cmd_nonblocking=0;
lp->exec_box->mbox=0;
lp->exec_box->mbox=cmd;
memcpy((void *)lp->exec_box->data, data, len);
barrier(); /* the memcpy forgot the volatile so be sure */
mc32_ready_poll(dev);
outb(1<<6, ioaddr+HOST_CMD);
wait_for_completion(&lp->execution_cmd);
if(lp->exec_box->mbox&(1<<13))
ret = -1;
up(&lp->cmd_mutex);
/*
* A multicast set got blocked - try it now
*/
if(lp->mc_reload_wait)
{
mc32_reset_multicast_list(dev);
}
return ret;
}
/**
* mc32_start_transceiver - tell board to restart tx/rx
* @dev: The 3c527 card to issue the command to
*
* This may be called from the interrupt state, where it is used
* to restart the rx ring if the card runs out of rx buffers.
*
* We must first check if it's ok to (re)start the transceiver. See
* mc32_close for details.
*/
static void mc32_start_transceiver(struct net_device *dev) {
struct mc32_local *lp = netdev_priv(dev);
int ioaddr = dev->base_addr;
/* Ignore RX overflow on device closure */
if (lp->xceiver_desired_state==HALTED)
return;
/* Give the card the offset to the post-EOL-bit RX descriptor */
mc32_ready_poll(dev);
lp->rx_box->mbox=0;
lp->rx_box->data[0]=lp->rx_ring[prev_rx(lp->rx_ring_tail)].p->next;
outb(HOST_CMD_START_RX, ioaddr+HOST_CMD);
mc32_ready_poll(dev);
lp->tx_box->mbox=0;
outb(HOST_CMD_RESTRT_TX, ioaddr+HOST_CMD); /* card ignores this on RX restart */
/* We are not interrupted on start completion */
}
/**
* mc32_halt_transceiver - tell board to stop tx/rx
* @dev: The 3c527 card to issue the command to
*
* We issue the commands to halt the card's transceiver. In fact,
* after some experimenting we now simply tell the card to
* suspend. When issuing aborts occasionally odd things happened.
*
* We then sleep until the card has notified us that both rx and
* tx have been suspended.
*/
static void mc32_halt_transceiver(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
int ioaddr = dev->base_addr;
mc32_ready_poll(dev);
lp->rx_box->mbox=0;
outb(HOST_CMD_SUSPND_RX, ioaddr+HOST_CMD);
wait_for_completion(&lp->xceiver_cmd);
mc32_ready_poll(dev);
lp->tx_box->mbox=0;
outb(HOST_CMD_SUSPND_TX, ioaddr+HOST_CMD);
wait_for_completion(&lp->xceiver_cmd);
}
/**
* mc32_load_rx_ring - load the ring of receive buffers
* @dev: 3c527 to build the ring for
*
* This initalises the on-card and driver datastructures to
* the point where mc32_start_transceiver() can be called.
*
* The card sets up the receive ring for us. We are required to use the
* ring it provides, although the size of the ring is configurable.
*
* We allocate an sk_buff for each ring entry in turn and
* initalise its house-keeping info. At the same time, we read
* each 'next' pointer in our rx_ring array. This reduces slow
* shared-memory reads and makes it easy to access predecessor
* descriptors.
*
* We then set the end-of-list bit for the last entry so that the
* card will know when it has run out of buffers.
*/
static int mc32_load_rx_ring(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
int i;
u16 rx_base;
volatile struct skb_header *p;
rx_base=lp->rx_chain;
for(i=0; i<RX_RING_LEN; i++) {
lp->rx_ring[i].skb=alloc_skb(1532, GFP_KERNEL);
if (lp->rx_ring[i].skb==NULL) {
for (;i>=0;i--)
kfree_skb(lp->rx_ring[i].skb);
return -ENOBUFS;
}
skb_reserve(lp->rx_ring[i].skb, 18);
p=isa_bus_to_virt(lp->base+rx_base);
p->control=0;
p->data=isa_virt_to_bus(lp->rx_ring[i].skb->data);
p->status=0;
p->length=1532;
lp->rx_ring[i].p=p;
rx_base=p->next;
}
lp->rx_ring[i-1].p->control |= CONTROL_EOL;
lp->rx_ring_tail=0;
return 0;
}
/**
* mc32_flush_rx_ring - free the ring of receive buffers
* @lp: Local data of 3c527 to flush the rx ring of
*
* Free the buffer for each ring slot. This may be called
* before mc32_load_rx_ring(), eg. on error in mc32_open().
* Requires rx skb pointers to point to a valid skb, or NULL.
*/
static void mc32_flush_rx_ring(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
int i;
for(i=0; i < RX_RING_LEN; i++)
{
if (lp->rx_ring[i].skb) {
dev_kfree_skb(lp->rx_ring[i].skb);
lp->rx_ring[i].skb = NULL;
}
lp->rx_ring[i].p=NULL;
}
}
/**
* mc32_load_tx_ring - load transmit ring
* @dev: The 3c527 card to issue the command to
*
* This sets up the host transmit data-structures.
*
* First, we obtain from the card it's current postion in the tx
* ring, so that we will know where to begin transmitting
* packets.
*
* Then, we read the 'next' pointers from the on-card tx ring into
* our tx_ring array to reduce slow shared-mem reads. Finally, we
* intitalise the tx house keeping variables.
*
*/
static void mc32_load_tx_ring(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
volatile struct skb_header *p;
int i;
u16 tx_base;
tx_base=lp->tx_box->data[0];
for(i=0 ; i<TX_RING_LEN ; i++)
{
p=isa_bus_to_virt(lp->base+tx_base);
lp->tx_ring[i].p=p;
lp->tx_ring[i].skb=NULL;
tx_base=p->next;
}
/* -1 so that tx_ring_head cannot "lap" tx_ring_tail */
/* see mc32_tx_ring */
atomic_set(&lp->tx_count, TX_RING_LEN-1);
atomic_set(&lp->tx_ring_head, 0);
lp->tx_ring_tail=0;
}
/**
* mc32_flush_tx_ring - free transmit ring
* @lp: Local data of 3c527 to flush the tx ring of
*
* If the ring is non-empty, zip over the it, freeing any
* allocated skb_buffs. The tx ring house-keeping variables are
* then reset. Requires rx skb pointers to point to a valid skb,
* or NULL.
*/
static void mc32_flush_tx_ring(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
int i;
for (i=0; i < TX_RING_LEN; i++)
{
if (lp->tx_ring[i].skb)
{
dev_kfree_skb(lp->tx_ring[i].skb);
lp->tx_ring[i].skb = NULL;
}
}
atomic_set(&lp->tx_count, 0);
atomic_set(&lp->tx_ring_head, 0);
lp->tx_ring_tail=0;
}
/**
* mc32_open - handle 'up' of card
* @dev: device to open
*
* The user is trying to bring the card into ready state. This requires
* a brief dialogue with the card. Firstly we enable interrupts and then
* 'indications'. Without these enabled the card doesn't bother telling
* us what it has done. This had me puzzled for a week.
*
* We configure the number of card descriptors, then load the network
* address and multicast filters. Turn on the workaround mode. This
* works around a bug in the 82586 - it asks the firmware to do
* so. It has a performance (latency) hit but is needed on busy
* [read most] lans. We load the ring with buffers then we kick it
* all off.
*/
static int mc32_open(struct net_device *dev)
{
int ioaddr = dev->base_addr;
struct mc32_local *lp = netdev_priv(dev);
u8 one=1;
u8 regs;
u16 descnumbuffs[2] = {TX_RING_LEN, RX_RING_LEN};
/*
* Interrupts enabled
*/
regs=inb(ioaddr+HOST_CTRL);
regs|=HOST_CTRL_INTE;
outb(regs, ioaddr+HOST_CTRL);
/*
* Allow ourselves to issue commands
*/
up(&lp->cmd_mutex);
/*
* Send the indications on command
*/
mc32_command(dev, 4, &one, 2);
/*
* Poke it to make sure it's really dead.
*/
mc32_halt_transceiver(dev);
mc32_flush_tx_ring(dev);
/*
* Ask card to set up on-card descriptors to our spec
*/
if(mc32_command(dev, 8, descnumbuffs, 4)) {
printk("%s: %s rejected our buffer configuration!\n",
dev->name, cardname);
mc32_close(dev);
return -ENOBUFS;
}
/* Report new configuration */
mc32_command(dev, 6, NULL, 0);
lp->tx_chain = lp->exec_box->data[8]; /* Transmit list start offset */
lp->rx_chain = lp->exec_box->data[10]; /* Receive list start offset */
lp->tx_len = lp->exec_box->data[9]; /* Transmit list count */
lp->rx_len = lp->exec_box->data[11]; /* Receive list count */
/* Set Network Address */
mc32_command(dev, 1, dev->dev_addr, 6);
/* Set the filters */
mc32_set_multicast_list(dev);
if (WORKAROUND_82586) {
u16 zero_word=0;
mc32_command(dev, 0x0D, &zero_word, 2); /* 82586 bug workaround on */
}
mc32_load_tx_ring(dev);
if(mc32_load_rx_ring(dev))
{
mc32_close(dev);
return -ENOBUFS;
}
lp->xceiver_desired_state = RUNNING;
/* And finally, set the ball rolling... */
mc32_start_transceiver(dev);
netif_start_queue(dev);
return 0;
}
/**
* mc32_timeout - handle a timeout from the network layer
* @dev: 3c527 that timed out
*
* Handle a timeout on transmit from the 3c527. This normally means
* bad things as the hardware handles cable timeouts and mess for
* us.
*
*/
static void mc32_timeout(struct net_device *dev)
{
printk(KERN_WARNING "%s: transmit timed out?\n", dev->name);
/* Try to restart the adaptor. */
netif_wake_queue(dev);
}
/**
* mc32_send_packet - queue a frame for transmit
* @skb: buffer to transmit
* @dev: 3c527 to send it out of
*
* Transmit a buffer. This normally means throwing the buffer onto
* the transmit queue as the queue is quite large. If the queue is
* full then we set tx_busy and return. Once the interrupt handler
* gets messages telling it to reclaim transmit queue entries, we will
* clear tx_busy and the kernel will start calling this again.
*
* We do not disable interrupts or acquire any locks; this can
* run concurrently with mc32_tx_ring(), and the function itself
* is serialised at a higher layer. However, similarly for the
* card itself, we must ensure that we update tx_ring_head only
* after we've established a valid packet on the tx ring (and
* before we let the card "see" it, to prevent it racing with the
* irq handler).
*
*/
static int mc32_send_packet(struct sk_buff *skb, struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
u32 head = atomic_read(&lp->tx_ring_head);
volatile struct skb_header *p, *np;
netif_stop_queue(dev);
if(atomic_read(&lp->tx_count)==0) {
return 1;
}
skb = skb_padto(skb, ETH_ZLEN);
if (skb == NULL) {
netif_wake_queue(dev);
return 0;
}
atomic_dec(&lp->tx_count);
/* P is the last sending/sent buffer as a pointer */
p=lp->tx_ring[head].p;
head = next_tx(head);
/* NP is the buffer we will be loading */
np=lp->tx_ring[head].p;
/* We will need this to flush the buffer out */
lp->tx_ring[head].skb=skb;
np->length = unlikely(skb->len < ETH_ZLEN) ? ETH_ZLEN : skb->len;
np->data = isa_virt_to_bus(skb->data);
np->status = 0;
np->control = CONTROL_EOP | CONTROL_EOL;
wmb();
/*
* The new frame has been setup; we can now
* let the interrupt handler and card "see" it
*/
atomic_set(&lp->tx_ring_head, head);
p->control &= ~CONTROL_EOL;
netif_wake_queue(dev);
return 0;
}
/**
* mc32_update_stats - pull off the on board statistics
* @dev: 3c527 to service
*
*
* Query and reset the on-card stats. There's the small possibility
* of a race here, which would result in an underestimation of
* actual errors. As such, we'd prefer to keep all our stats
* collection in software. As a rule, we do. However it can't be
* used for rx errors and collisions as, by default, the card discards
* bad rx packets.
*
* Setting the SAV BP in the rx filter command supposedly
* stops this behaviour. However, testing shows that it only seems to
* enable the collation of on-card rx statistics --- the driver
* never sees an RX descriptor with an error status set.
*
*/
static void mc32_update_stats(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
volatile struct mc32_stats *st = lp->stats;
u32 rx_errors=0;
rx_errors+=lp->net_stats.rx_crc_errors +=st->rx_crc_errors;
st->rx_crc_errors=0;
rx_errors+=lp->net_stats.rx_fifo_errors +=st->rx_overrun_errors;
st->rx_overrun_errors=0;
rx_errors+=lp->net_stats.rx_frame_errors +=st->rx_alignment_errors;
st->rx_alignment_errors=0;
rx_errors+=lp->net_stats.rx_length_errors+=st->rx_tooshort_errors;
st->rx_tooshort_errors=0;
rx_errors+=lp->net_stats.rx_missed_errors+=st->rx_outofresource_errors;
st->rx_outofresource_errors=0;
lp->net_stats.rx_errors=rx_errors;
/* Number of packets which saw one collision */
lp->net_stats.collisions+=st->dataC[10];
st->dataC[10]=0;
/* Number of packets which saw 2--15 collisions */
lp->net_stats.collisions+=st->dataC[11];
st->dataC[11]=0;
}
/**
* mc32_rx_ring - process the receive ring
* @dev: 3c527 that needs its receive ring processing
*
*
* We have received one or more indications from the card that a
* receive has completed. The buffer ring thus contains dirty
* entries. We walk the ring by iterating over the circular rx_ring
* array, starting at the next dirty buffer (which happens to be the
* one we finished up at last time around).
*
* For each completed packet, we will either copy it and pass it up
* the stack or, if the packet is near MTU sized, we allocate
* another buffer and flip the old one up the stack.
*
* We must succeed in keeping a buffer on the ring. If necessary we
* will toss a received packet rather than lose a ring entry. Once
* the first uncompleted descriptor is found, we move the
* End-Of-List bit to include the buffers just processed.
*
*/
static void mc32_rx_ring(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
volatile struct skb_header *p;
u16 rx_ring_tail;
u16 rx_old_tail;
int x=0;
rx_old_tail = rx_ring_tail = lp->rx_ring_tail;
do
{
p=lp->rx_ring[rx_ring_tail].p;
if(!(p->status & (1<<7))) { /* Not COMPLETED */
break;
}
if(p->status & (1<<6)) /* COMPLETED_OK */
{
u16 length=p->length;
struct sk_buff *skb;
struct sk_buff *newskb;
/* Try to save time by avoiding a copy on big frames */
if ((length > RX_COPYBREAK)
&& ((newskb=dev_alloc_skb(1532)) != NULL))
{
skb=lp->rx_ring[rx_ring_tail].skb;
skb_put(skb, length);
skb_reserve(newskb,18);
lp->rx_ring[rx_ring_tail].skb=newskb;
p->data=isa_virt_to_bus(newskb->data);
}
else
{
skb=dev_alloc_skb(length+2);
if(skb==NULL) {
lp->net_stats.rx_dropped++;
goto dropped;
}
skb_reserve(skb,2);
memcpy(skb_put(skb, length),
lp->rx_ring[rx_ring_tail].skb->data, length);
}
skb->protocol=eth_type_trans(skb,dev);
skb->dev=dev;
dev->last_rx = jiffies;
lp->net_stats.rx_packets++;
lp->net_stats.rx_bytes += length;
netif_rx(skb);
}
dropped:
p->length = 1532;
p->status = 0;
rx_ring_tail=next_rx(rx_ring_tail);
}
while(x++<48);
/* If there was actually a frame to be processed, place the EOL bit */
/* at the descriptor prior to the one to be filled next */
if (rx_ring_tail != rx_old_tail)
{
lp->rx_ring[prev_rx(rx_ring_tail)].p->control |= CONTROL_EOL;
lp->rx_ring[prev_rx(rx_old_tail)].p->control &= ~CONTROL_EOL;
lp->rx_ring_tail=rx_ring_tail;
}
}
/**
* mc32_tx_ring - process completed transmits
* @dev: 3c527 that needs its transmit ring processing
*
*
* This operates in a similar fashion to mc32_rx_ring. We iterate
* over the transmit ring. For each descriptor which has been
* processed by the card, we free its associated buffer and note
* any errors. This continues until the transmit ring is emptied
* or we reach a descriptor that hasn't yet been processed by the
* card.
*
*/
static void mc32_tx_ring(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
volatile struct skb_header *np;
/*
* We rely on head==tail to mean 'queue empty'.
* This is why lp->tx_count=TX_RING_LEN-1: in order to prevent
* tx_ring_head wrapping to tail and confusing a 'queue empty'
* condition with 'queue full'
*/
while (lp->tx_ring_tail != atomic_read(&lp->tx_ring_head))
{
u16 t;
t=next_tx(lp->tx_ring_tail);
np=lp->tx_ring[t].p;
if(!(np->status & (1<<7)))
{
/* Not COMPLETED */
break;
}
lp->net_stats.tx_packets++;
if(!(np->status & (1<<6))) /* Not COMPLETED_OK */
{
lp->net_stats.tx_errors++;
switch(np->status&0x0F)
{
case 1:
lp->net_stats.tx_aborted_errors++;
break; /* Max collisions */
case 2:
lp->net_stats.tx_fifo_errors++;
break;
case 3:
lp->net_stats.tx_carrier_errors++;
break;
case 4:
lp->net_stats.tx_window_errors++;
break; /* CTS Lost */
case 5:
lp->net_stats.tx_aborted_errors++;
break; /* Transmit timeout */
}
}
/* Packets are sent in order - this is
basically a FIFO queue of buffers matching
the card ring */
lp->net_stats.tx_bytes+=lp->tx_ring[t].skb->len;
dev_kfree_skb_irq(lp->tx_ring[t].skb);
lp->tx_ring[t].skb=NULL;
atomic_inc(&lp->tx_count);
netif_wake_queue(dev);
lp->tx_ring_tail=t;
}
}
/**
* mc32_interrupt - handle an interrupt from a 3c527
* @irq: Interrupt number
* @dev_id: 3c527 that requires servicing
* @regs: Registers (unused)
*
*
* An interrupt is raised whenever the 3c527 writes to the command
* register. This register contains the message it wishes to send us
* packed into a single byte field. We keep reading status entries
* until we have processed all the control items, but simply count
* transmit and receive reports. When all reports are in we empty the
* transceiver rings as appropriate. This saves the overhead of
* multiple command requests.
*
* Because MCA is level-triggered, we shouldn't miss indications.
* Therefore, we needn't ask the card to suspend interrupts within
* this handler. The card receives an implicit acknowledgment of the
* current interrupt when we read the command register.
*
*/
static irqreturn_t mc32_interrupt(int irq, void *dev_id, struct pt_regs * regs)
{
struct net_device *dev = dev_id;
struct mc32_local *lp;
int ioaddr, status, boguscount = 0;
int rx_event = 0;
int tx_event = 0;
if (dev == NULL) {
printk(KERN_WARNING "%s: irq %d for unknown device.\n", cardname, irq);
return IRQ_NONE;
}
ioaddr = dev->base_addr;
lp = netdev_priv(dev);
/* See whats cooking */
while((inb(ioaddr+HOST_STATUS)&HOST_STATUS_CWR) && boguscount++<2000)
{
status=inb(ioaddr+HOST_CMD);
#ifdef DEBUG_IRQ
printk("Status TX%d RX%d EX%d OV%d BC%d\n",
(status&7), (status>>3)&7, (status>>6)&1,
(status>>7)&1, boguscount);
#endif
switch(status&7)
{
case 0:
break;
case 6: /* TX fail */
case 2: /* TX ok */
tx_event = 1;
break;
case 3: /* Halt */
case 4: /* Abort */
complete(&lp->xceiver_cmd);
break;
default:
printk("%s: strange tx ack %d\n", dev->name, status&7);
}
status>>=3;
switch(status&7)
{
case 0:
break;
case 2: /* RX */
rx_event=1;
break;
case 3: /* Halt */
case 4: /* Abort */
complete(&lp->xceiver_cmd);
break;
case 6:
/* Out of RX buffers stat */
/* Must restart rx */
lp->net_stats.rx_dropped++;
mc32_rx_ring(dev);
mc32_start_transceiver(dev);
break;
default:
printk("%s: strange rx ack %d\n",
dev->name, status&7);
}
status>>=3;
if(status&1)
{
/*
* No thread is waiting: we need to tidy
* up ourself.
*/
if (lp->cmd_nonblocking) {
up(&lp->cmd_mutex);
if (lp->mc_reload_wait)
mc32_reset_multicast_list(dev);
}
else complete(&lp->execution_cmd);
}
if(status&2)
{
/*
* We get interrupted once per
* counter that is about to overflow.
*/
mc32_update_stats(dev);
}
}
/*
* Process the transmit and receive rings
*/
if(tx_event)
mc32_tx_ring(dev);
if(rx_event)
mc32_rx_ring(dev);
return IRQ_HANDLED;
}
/**
* mc32_close - user configuring the 3c527 down
* @dev: 3c527 card to shut down
*
* The 3c527 is a bus mastering device. We must be careful how we
* shut it down. It may also be running shared interrupt so we have
* to be sure to silence it properly
*
* We indicate that the card is closing to the rest of the
* driver. Otherwise, it is possible that the card may run out
* of receive buffers and restart the transceiver while we're
* trying to close it.
*
* We abort any receive and transmits going on and then wait until
* any pending exec commands have completed in other code threads.
* In theory we can't get here while that is true, in practice I am
* paranoid
*
* We turn off the interrupt enable for the board to be sure it can't
* intefere with other devices.
*/
static int mc32_close(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
int ioaddr = dev->base_addr;
u8 regs;
u16 one=1;
lp->xceiver_desired_state = HALTED;
netif_stop_queue(dev);
/*
* Send the indications on command (handy debug check)
*/
mc32_command(dev, 4, &one, 2);
/* Shut down the transceiver */
mc32_halt_transceiver(dev);
/* Ensure we issue no more commands beyond this point */
down(&lp->cmd_mutex);
/* Ok the card is now stopping */
regs=inb(ioaddr+HOST_CTRL);
regs&=~HOST_CTRL_INTE;
outb(regs, ioaddr+HOST_CTRL);
mc32_flush_rx_ring(dev);
mc32_flush_tx_ring(dev);
mc32_update_stats(dev);
return 0;
}
/**
* mc32_get_stats - hand back stats to network layer
* @dev: The 3c527 card to handle
*
* We've collected all the stats we can in software already. Now
* it's time to update those kept on-card and return the lot.
*
*/
static struct net_device_stats *mc32_get_stats(struct net_device *dev)
{
struct mc32_local *lp = netdev_priv(dev);
mc32_update_stats(dev);
return &lp->net_stats;
}
/**
* do_mc32_set_multicast_list - attempt to update multicasts
* @dev: 3c527 device to load the list on
* @retry: indicates this is not the first call.
*
*
* Actually set or clear the multicast filter for this adaptor. The
* locking issues are handled by this routine. We have to track
* state as it may take multiple calls to get the command sequence
* completed. We just keep trying to schedule the loads until we
* manage to process them all.
*
* 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.
*
* See mc32_update_stats() regards setting the SAV BP bit.
*
*/
static void do_mc32_set_multicast_list(struct net_device *dev, int retry)
{
struct mc32_local *lp = netdev_priv(dev);
u16 filt = (1<<2); /* Save Bad Packets, for stats purposes */
if (dev->flags&IFF_PROMISC)
/* Enable promiscuous mode */
filt |= 1;
else if((dev->flags&IFF_ALLMULTI) || dev->mc_count > 10)
{
dev->flags|=IFF_PROMISC;
filt |= 1;
}
else if(dev->mc_count)
{
unsigned char block[62];
unsigned char *bp;
struct dev_mc_list *dmc=dev->mc_list;
int i;
if(retry==0)
lp->mc_list_valid = 0;
if(!lp->mc_list_valid)
{
block[1]=0;
block[0]=dev->mc_count;
bp=block+2;
for(i=0;i<dev->mc_count;i++)
{
memcpy(bp, dmc->dmi_addr, 6);
bp+=6;
dmc=dmc->next;
}
if(mc32_command_nowait(dev, 2, block, 2+6*dev->mc_count)==-1)
{
lp->mc_reload_wait = 1;
return;
}
lp->mc_list_valid=1;
}
}
if(mc32_command_nowait(dev, 0, &filt, 2)==-1)
{
lp->mc_reload_wait = 1;
}
else {
lp->mc_reload_wait = 0;
}
}
/**
* mc32_set_multicast_list - queue multicast list update
* @dev: The 3c527 to use
*
* Commence loading the multicast list. This is called when the kernel
* changes the lists. It will override any pending list we are trying to
* load.
*/
static void mc32_set_multicast_list(struct net_device *dev)
{
do_mc32_set_multicast_list(dev,0);
}
/**
* mc32_reset_multicast_list - reset multicast list
* @dev: The 3c527 to use
*
* Attempt the next step in loading the multicast lists. If this attempt
* fails to complete then it will be scheduled and this function called
* again later from elsewhere.
*/
static void mc32_reset_multicast_list(struct net_device *dev)
{
do_mc32_set_multicast_list(dev,1);
}
static void netdev_get_drvinfo(struct net_device *dev,
struct ethtool_drvinfo *info)
{
strcpy(info->driver, DRV_NAME);
strcpy(info->version, DRV_VERSION);
sprintf(info->bus_info, "MCA 0x%lx", dev->base_addr);
}
static u32 netdev_get_msglevel(struct net_device *dev)
{
return mc32_debug;
}
static void netdev_set_msglevel(struct net_device *dev, u32 level)
{
mc32_debug = level;
}
static struct ethtool_ops netdev_ethtool_ops = {
.get_drvinfo = netdev_get_drvinfo,
.get_msglevel = netdev_get_msglevel,
.set_msglevel = netdev_set_msglevel,
};
#ifdef MODULE
static struct net_device *this_device;
/**
* init_module - entry point
*
* Probe and locate a 3c527 card. This really should probe and locate
* all the 3c527 cards in the machine not just one of them. Yes you can
* insmod multiple modules for now but it's a hack.
*/
int init_module(void)
{
this_device = mc32_probe(-1);
if (IS_ERR(this_device))
return PTR_ERR(this_device);
return 0;
}
/**
* cleanup_module - free resources for an unload
*
* Unloading time. We release the MCA bus resources and the interrupt
* at which point everything is ready to unload. The card must be stopped
* at this point or we would not have been called. When we unload we
* leave the card stopped but not totally shut down. When the card is
* initialized it must be rebooted or the rings reloaded before any
* transmit operations are allowed to start scribbling into memory.
*/
void cleanup_module(void)
{
unregister_netdev(this_device);
cleanup_card(this_device);
free_netdev(this_device);
}
#endif /* MODULE */