hi
this patch does:
- net_device is no longer allocated as part of the driver's private structure,
instead it's allocated via alloc_netdev
- xirc2ps_detach calls xirc2ps_release if necessary (like the other drivers)
against 2.5.70-bk.
rgds
-daniel
===== drivers/net/pcmcia/xirc2ps_cs.c 1.19 vs edited =====
--- 1.19/drivers/net/pcmcia/xirc2ps_cs.c Sun May 25 17:07:51 2003
+++ edited/drivers/net/pcmcia/xirc2ps_cs.c Mon Jun 9 15:27:53 2003
@@ -358,7 +358,6 @@
typedef struct local_info_t {
dev_link_t link;
- struct net_device dev;
dev_node_t node;
struct net_device_stats stats;
int card_type;
@@ -619,11 +618,12 @@
flush_stale_links();
/* Allocate the device structure */
- local = kmalloc(sizeof(*local), GFP_KERNEL);
- if (!local) return NULL;
- memset(local, 0, sizeof(*local));
- link = &local->link; dev = &local->dev;
- link->priv = dev->priv = local;
+ dev = alloc_etherdev(sizeof(local_info_t));
+ if (!dev)
+ return NULL;
+ local = dev->priv;
+ link = &local->link;
+ link->priv = dev;
init_timer(&link->release);
link->release.function = &xirc2ps_release;
@@ -645,7 +645,6 @@
dev->get_stats = &do_get_stats;
dev->do_ioctl = &do_ioctl;
dev->set_multicast_list = &set_multicast_list;
- ether_setup(dev);
dev->open = &do_open;
dev->stop = &do_stop;
#ifdef HAVE_TX_TIMEOUT
@@ -684,7 +683,7 @@
static void
xirc2ps_detach(dev_link_t * link)
{
- local_info_t *local = link->priv;
+ struct net_device *dev = link->priv;
dev_link_t **linkp;
DEBUG(0, "detach(0x%p)\n", link);
@@ -706,10 +705,11 @@
*/
del_timer(&link->release);
if (link->state & DEV_CONFIG) {
- DEBUG(0, "detach postponed, '%s' still locked\n",
- link->dev->dev_name);
- link->state |= DEV_STALE_LINK;
- return;
+ xirc2ps_release((unsigned long)link);
+ if (link->state & DEV_STALE_CONFIG) {
+ link->state |= DEV_STALE_LINK;
+ return;
+ }
}
/* Break the link with Card Services */
@@ -719,8 +719,8 @@
/* Unlink device structure, free it */
*linkp = link->next;
if (link->dev)
- unregister_netdev(&local->dev);
- kfree(local);
+ unregister_netdev(dev);
+ kfree(dev);
} /* xirc2ps_detach */
@@ -745,7 +745,8 @@
static int
set_card_type(dev_link_t *link, const void *s)
{
- local_info_t *local = link->priv;
+ struct net_device *dev = link->priv;
+ local_info_t *local = dev->priv;
#ifdef PCMCIA_DEBUG
unsigned cisrev = ((const unsigned char *)s)[2];
#endif
@@ -839,8 +840,8 @@
xirc2ps_config(dev_link_t * link)
{
client_handle_t handle = link->handle;
- local_info_t *local = link->priv;
- struct net_device *dev = &local->dev;
+ struct net_device *dev = link->priv;
+ local_info_t *local = dev->priv;
tuple_t tuple;
cisparse_t parse;
ioaddr_t ioaddr;
@@ -1195,11 +1196,10 @@
xirc2ps_release(u_long arg)
{
dev_link_t *link = (dev_link_t *) arg;
- local_info_t *local = link->priv;
- struct net_device *dev = &local->dev;
DEBUG(0, "release(0x%p)\n", link);
+#if 0
/*
* If the device is currently in use, we won't release until it
* is actually closed.
@@ -1210,8 +1210,10 @@
link->state |= DEV_STALE_CONFIG;
return;
}
+#endif
if (link->win) {
+ struct net_device *dev = link->priv;
local_info_t *local = dev->priv;
if (local->dingo)
iounmap(local->dingo_ccr - 0x0800);
@@ -1243,8 +1245,7 @@
event_callback_args_t * args)
{
dev_link_t *link = args->client_data;
- local_info_t *lp = link->priv;
- struct net_device *dev = &lp->dev;
+ struct net_device *dev = link->priv;
DEBUG(0, "event(%d)\n", (int)event);
@@ -2083,12 +2084,8 @@
{
pcmcia_unregister_driver(&xirc2ps_cs_driver);
- while (dev_list) {
- if (dev_list->state & DEV_CONFIG)
- xirc2ps_release((u_long)dev_list);
- if (dev_list) /* xirc2ps_release() might already have detached... */
- xirc2ps_detach(dev_list);
- }
+ while (dev_list)
+ xirc2ps_detach(dev_list);
}
module_init(init_xirc2ps_cs);
the second patch:
replaces busy_loop with a simple macro doing a schedule_timeout. busy_loop was never
called from interrupt conext anyway, so no need for that. and the sti() is gone.
rgds
-daniel
--- linux-2.5/drivers/net/pcmcia/xirc2ps_cs.c~1 2003-06-09 15:28:22.000000000 -0400
+++ linux-2.5/drivers/net/pcmcia/xirc2ps_cs.c 2003-06-10 14:17:04.000000000 -0400
@@ -431,22 +431,10 @@
#define PutByte(reg,value) outb((value), ioaddr+(reg))
#define PutWord(reg,value) outw((value), ioaddr+(reg))
-static void
-busy_loop(u_long len)
-{
- if (in_interrupt()) {
- u_long timeout = jiffies + len;
- u_long flags;
- save_flags(flags);
- sti();
- while (time_before_eq(jiffies, timeout))
- ;
- restore_flags(flags);
- } else {
- __set_current_state(TASK_UNINTERRUPTIBLE);
- schedule_timeout(len);
- }
-}
+#define Wait(n) do { \
+ set_current_state(TASK_UNINTERRUPTIBLE); \
+ schedule_timeout(n); \
+} while (0)
/*====== Functions used for debugging =================================*/
#if defined(PCMCIA_DEBUG) && 0 /* reading regs may change system status */
@@ -1780,12 +1768,12 @@
SelectPage(4);
udelay(1);
PutByte(XIRCREG4_GPR1, 0); /* clear bit 0: power down */
- busy_loop(HZ/25); /* wait 40 msec */
+ Wait(HZ/25); /* wait 40 msec */
if (local->mohawk)
PutByte(XIRCREG4_GPR1, 1); /* set bit 0: power up */
else
PutByte(XIRCREG4_GPR1, 1 | 4); /* set bit 0: power up, bit 2: AIC */
- busy_loop(HZ/50); /* wait 20 msec */
+ Wait(HZ/50); /* wait 20 msec */
}
static void
@@ -1799,9 +1787,9 @@
hardreset(dev);
PutByte(XIRCREG_CR, SoftReset); /* set */
- busy_loop(HZ/50); /* wait 20 msec */
+ Wait(HZ/50); /* wait 20 msec */
PutByte(XIRCREG_CR, 0); /* clear */
- busy_loop(HZ/25); /* wait 40 msec */
+ Wait(HZ/25); /* wait 40 msec */
if (local->mohawk) {
SelectPage(4);
/* set pin GP1 and GP2 to output (0x0c)
@@ -1812,7 +1800,7 @@
}
/* give the circuits some time to power up */
- busy_loop(HZ/2); /* about 500ms */
+ Wait(HZ/2); /* about 500ms */
local->last_ptr_value = 0;
local->silicon = local->mohawk ? (GetByte(XIRCREG4_BOV) & 0x70) >> 4
@@ -1831,7 +1819,7 @@
SelectPage(0x42);
PutByte(XIRCREG42_SWC1, 0x80);
}
- busy_loop(HZ/25); /* wait 40 msec to let it complete */
+ Wait(HZ/25); /* wait 40 msec to let it complete */
#ifdef PCMCIA_DEBUG
if (pc_debug) {
@@ -1890,7 +1878,7 @@
printk(KERN_INFO "%s: MII selected\n", dev->name);
SelectPage(2);
PutByte(XIRCREG2_MSR, GetByte(XIRCREG2_MSR) | 0x08);
- busy_loop(HZ/50);
+ Wait(HZ/50);
} else {
printk(KERN_INFO "%s: MII detected; using 10mbs\n",
dev->name);
@@ -1899,7 +1887,7 @@
PutByte(XIRCREG42_SWC1, 0xC0);
else /* enable 10BaseT */
PutByte(XIRCREG42_SWC1, 0x80);
- busy_loop(HZ/25); /* wait 40 msec to let it complete */
+ Wait(HZ/25); /* wait 40 msec to let it complete */
}
if (full_duplex)
PutByte(XIRCREG1_ECR, GetByte(XIRCREG1_ECR | FullDuplex));
@@ -1992,7 +1980,7 @@
* Fixme: Better to use a timer here!
*/
for (i=0; i < 35; i++) {
- busy_loop(HZ/10); /* wait 100 msec */
+ Wait(HZ/10); /* wait 100 msec */
status = mii_rd(ioaddr, 0, 1);
if ((status & 0x0020) && (status & 0x0004))
break;
On Tue, 10 Jun 2003, Daniel Ritz wrote:
> - busy_loop(HZ/25); /* wait 40 msec */
> + Wait(HZ/25); /* wait 40 msec */
Why not Wait(40); instead Wait(HZ/25) ? Currently HZ is 1000. However, the
value can change - as it changed from 100 to 1000.
Geller Sandor <[email protected]>
On Tue, Jun 10, 2003 at 05:35:18PM +0200, Geller Sandor wrote:
> On Tue, 10 Jun 2003, Daniel Ritz wrote:
>
> > - busy_loop(HZ/25); /* wait 40 msec */
> > + Wait(HZ/25); /* wait 40 msec */
>
> Why not Wait(40); instead Wait(HZ/25) ? Currently HZ is 1000. However, the
> value can change - as it changed from 100 to 1000.
True enough... the best solution is to grep the tree for a
msecs_to_jiffies macro, and use that. Then it will look like
Wait(msecs_to_jiffies(40))
and the macro does the proper scaling versus constant HZ.
Jeff
On Tuesday 10 June 2003 11:46, Jeff Garzik wrote:
> On Tue, Jun 10, 2003 at 05:35:18PM +0200, Geller Sandor wrote:
> > On Tue, 10 Jun 2003, Daniel Ritz wrote:
> > > - busy_loop(HZ/25); /* wait 40 msec */
> > > + Wait(HZ/25); /* wait 40 msec */
> >
> > Why not Wait(40); instead Wait(HZ/25) ? Currently HZ is 1000. However,
> > the value can change - as it changed from 100 to 1000.
>
> True enough... the best solution is to grep the tree for a
> msecs_to_jiffies macro, and use that. Then it will look like
>
hmm...yes, but the macro is defined in include/net/irda/irda.h
move it to a place where everyone can use it? like time.h or kernel.h?
> Wait(msecs_to_jiffies(40))
i would do it in the Wait macro. looks nicer..
and also define the Wait macro (with a better name, suggestions?)
somewhere else, 'cos other drivers use set_current_state(); schedule_timeout()
too..
>
> and the macro does the proper scaling versus constant HZ.
>
> Jeff
-daniel