OpenCloudOS-Kernel/drivers/net/appletalk/ltpc.c

1286 lines
31 KiB
C
Raw Normal View History

/*** ltpc.c -- a driver for the LocalTalk PC card.
*
* Copyright (c) 1995,1996 Bradford W. Johnson <johns393@maroon.tc.umn.edu>
*
* This software may be used and distributed according to the terms
* of the GNU General Public License, incorporated herein by reference.
*
* This is ALPHA code at best. It may not work for you. It may
* damage your equipment. It may damage your relations with other
* users of your network. Use it at your own risk!
*
* Based in part on:
* skeleton.c by Donald Becker
* dummy.c by Nick Holloway and Alan Cox
* loopback.c by Ross Biro, Fred van Kampen, Donald Becker
* the netatalk source code (UMICH)
* lots of work on the card...
*
* I do not have access to the (proprietary) SDK that goes with the card.
* If you do, I don't want to know about it, and you can probably write
* a better driver yourself anyway. This does mean that the pieces that
* talk to the card are guesswork on my part, so use at your own risk!
*
* This is my first try at writing Linux networking code, and is also
* guesswork. Again, use at your own risk! (Although on this part, I'd
* welcome suggestions)
*
* This is a loadable kernel module which seems to work at my site
* consisting of a 1.2.13 linux box running netatalk 1.3.3, and with
* the kernel support from 1.3.3b2 including patches routing.patch
* and ddp.disappears.from.chooser. In order to run it, you will need
* to patch ddp.c and aarp.c in the kernel, but only a little...
*
* I'm fairly confident that while this is arguably badly written, the
* problems that people experience will be "higher level", that is, with
* complications in the netatalk code. The driver itself doesn't do
* anything terribly complicated -- it pretends to be an ether device
* as far as netatalk is concerned, strips the DDP data out of the ether
* frame and builds a LLAP packet to send out the card. In the other
* direction, it receives LLAP frames from the card and builds a fake
* ether packet that it then tosses up to the networking code. You can
* argue (correctly) that this is an ugly way to do things, but it
* requires a minimal amount of fooling with the code in ddp.c and aarp.c.
*
* The card will do a lot more than is used here -- I *think* it has the
* layers up through ATP. Even if you knew how that part works (which I
* don't) it would be a big job to carve up the kernel ddp code to insert
* things at a higher level, and probably a bad idea...
*
* There are a number of other cards that do LocalTalk on the PC. If
* nobody finds any insurmountable (at the netatalk level) problems
* here, this driver should encourage people to put some work into the
* other cards (some of which I gather are still commercially available)
* and also to put hooks for LocalTalk into the official ddp code.
*
* I welcome comments and suggestions. This is my first try at Linux
* networking stuff, and there are probably lots of things that I did
* suboptimally.
*
***/
/***
*
* $Log: ltpc.c,v $
* Revision 1.1.2.1 2000/03/01 05:35:07 jgarzik
* at and tr cleanup
*
* Revision 1.8 1997/01/28 05:44:54 bradford
* Clean up for non-module a little.
* Hacked about a bit to clean things up - Alan Cox
* Probably broken it from the origina 1.8
*
* 1998/11/09: David Huggins-Daines <dhd@debian.org>
* Cleaned up the initialization code to use the standard autoirq methods,
and to probe for things in the standard order of i/o, irq, dma. This
removes the "reset the reset" hack, because I couldn't figure out an
easy way to get the card to trigger an interrupt after it.
* Added support for passing configuration parameters on the kernel command
line and through insmod
* Changed the device name from "ltalk0" to "lt0", both to conform with the
other localtalk driver, and to clear up the inconsistency between the
module and the non-module versions of the driver :-)
* Added a bunch of comments (I was going to make some enums for the state
codes and the register offsets, but I'm still not sure exactly what their
semantics are)
* Don't poll anymore in interrupt-driven mode
* It seems to work as a module now (as of 2.1.127), but I don't think
I'm responsible for that...
*
* Revision 1.7 1996/12/12 03:42:33 bradford
* DMA alloc cribbed from 3c505.c.
*
* Revision 1.6 1996/12/12 03:18:58 bradford
* Added virt_to_bus; works in 2.1.13.
*
* Revision 1.5 1996/12/12 03:13:22 root
* xmitQel initialization -- think through better though.
*
* Revision 1.4 1996/06/18 14:55:55 root
* Change names to ltpc. Tabs. Took a shot at dma alloc,
* although more needs to be done eventually.
*
* Revision 1.3 1996/05/22 14:59:39 root
* Change dev->open, dev->close to track dummy.c in 1.99.(around 7)
*
* Revision 1.2 1996/05/22 14:58:24 root
* Change tabs mostly.
*
* Revision 1.1 1996/04/23 04:45:09 root
* Initial revision
*
* Revision 0.16 1996/03/05 15:59:56 root
* Change ARPHRD_LOCALTLK definition to the "real" one.
*
* Revision 0.15 1996/03/05 06:28:30 root
* Changes for kernel 1.3.70. Still need a few patches to kernel, but
* it's getting closer.
*
* Revision 0.14 1996/02/25 17:38:32 root
* More cleanups. Removed query to card on get_stats.
*
* Revision 0.13 1996/02/21 16:27:40 root
* Refix debug_print_skb. Fix mac.raw gotcha that appeared in 1.3.65.
* Clean up receive code a little.
*
* Revision 0.12 1996/02/19 16:34:53 root
* Fix debug_print_skb. Kludge outgoing snet to 0 when using startup
* range. Change debug to mask: 1 for verbose, 2 for higher level stuff
* including packet printing, 4 for lower level (card i/o) stuff.
*
* Revision 0.11 1996/02/12 15:53:38 root
* Added router sends (requires new aarp.c patch)
*
* Revision 0.10 1996/02/11 00:19:35 root
* Change source LTALK_LOGGING debug switch to insmod ... debug=2.
*
* Revision 0.9 1996/02/10 23:59:35 root
* Fixed those fixes for 1.2 -- DANGER! The at.h that comes with netatalk
* has a *different* definition of struct sockaddr_at than the Linux kernel
* does. This is an "insidious and invidious" bug...
* (Actually the preceding comment is false -- it's the atalk.h in the
* ancient atalk-0.06 that's the problem)
*
* Revision 0.8 1996/02/10 19:09:00 root
* Merge 1.3 changes. Tested OK under 1.3.60.
*
* Revision 0.7 1996/02/10 17:56:56 root
* Added debug=1 parameter on insmod for debugging prints. Tried
* to fix timer unload on rmmod, but I don't think that's the problem.
*
* Revision 0.6 1995/12/31 19:01:09 root
* Clean up rmmod, irq comments per feedback from Corin Anderson (Thanks Corey!)
* Clean up initial probing -- sometimes the card wakes up latched in reset.
*
* Revision 0.5 1995/12/22 06:03:44 root
* Added comments in front and cleaned up a bit.
* This version sent out to people.
*
* Revision 0.4 1995/12/18 03:46:44 root
* Return shortDDP to longDDP fake to 0/0. Added command structs.
*
***/
/* ltpc jumpers are:
*
* Interrupts -- set at most one. If none are set, the driver uses
* polled mode. Because the card was developed in the XT era, the
* original documentation refers to IRQ2. Since you'll be running
* this on an AT (or later) class machine, that really means IRQ9.
*
* SW1 IRQ 4
* SW2 IRQ 3
* SW3 IRQ 9 (2 in original card documentation only applies to XT)
*
*
* DMA -- choose DMA 1 or 3, and set both corresponding switches.
*
* SW4 DMA 3
* SW5 DMA 1
* SW6 DMA 3
* SW7 DMA 1
*
*
* I/O address -- choose one.
*
* SW8 220 / 240
*/
/* To have some stuff logged, do
* insmod ltpc.o debug=1
*
* For a whole bunch of stuff, use higher numbers.
*
* The default is 0, i.e. no messages except for the probe results.
*/
/* insmod-tweakable variables */
static int debug;
#define DEBUG_VERBOSE 1
#define DEBUG_UPPER 2
#define DEBUG_LOWER 4
static int io;
static int irq;
static int dma;
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/types.h>
#include <linux/fcntl.h>
#include <linux/interrupt.h>
#include <linux/ptrace.h>
#include <linux/ioport.h>
#include <linux/spinlock.h>
#include <linux/in.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/if_arp.h>
#include <linux/if_ltalk.h>
#include <linux/delay.h>
#include <linux/timer.h>
#include <linux/atalk.h>
#include <linux/bitops.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 16:04:11 +08:00
#include <linux/gfp.h>
#include <asm/dma.h>
#include <asm/io.h>
/* our stuff */
#include "ltpc.h"
static DEFINE_SPINLOCK(txqueue_lock);
static DEFINE_SPINLOCK(mbox_lock);
/* function prototypes */
static int do_read(struct net_device *dev, void *cbuf, int cbuflen,
void *dbuf, int dbuflen);
static int sendup_buffer (struct net_device *dev);
/* Dma Memory related stuff, cribbed directly from 3c505.c */
static unsigned long dma_mem_alloc(int size)
{
int order = get_order(size);
return __get_dma_pages(GFP_KERNEL, order);
}
/* DMA data buffer, DMA command buffer */
static unsigned char *ltdmabuf;
static unsigned char *ltdmacbuf;
/* private struct, holds our appletalk address */
struct ltpc_private
{
struct atalk_addr my_addr;
};
/* transmit queue element struct */
struct xmitQel {
struct xmitQel *next;
/* command buffer */
unsigned char *cbuf;
short cbuflen;
/* data buffer */
unsigned char *dbuf;
short dbuflen;
unsigned char QWrite; /* read or write data */
unsigned char mailbox;
};
/* the transmit queue itself */
static struct xmitQel *xmQhd, *xmQtl;
static void enQ(struct xmitQel *qel)
{
unsigned long flags;
qel->next = NULL;
spin_lock_irqsave(&txqueue_lock, flags);
if (xmQtl) {
xmQtl->next = qel;
} else {
xmQhd = qel;
}
xmQtl = qel;
spin_unlock_irqrestore(&txqueue_lock, flags);
if (debug & DEBUG_LOWER)
printk("enqueued a 0x%02x command\n",qel->cbuf[0]);
}
static struct xmitQel *deQ(void)
{
unsigned long flags;
int i;
struct xmitQel *qel=NULL;
spin_lock_irqsave(&txqueue_lock, flags);
if (xmQhd) {
qel = xmQhd;
xmQhd = qel->next;
if(!xmQhd) xmQtl = NULL;
}
spin_unlock_irqrestore(&txqueue_lock, flags);
if ((debug & DEBUG_LOWER) && qel) {
int n;
printk(KERN_DEBUG "ltpc: dequeued command ");
n = qel->cbuflen;
if (n>100) n=100;
for(i=0;i<n;i++) printk("%02x ",qel->cbuf[i]);
printk("\n");
}
return qel;
}
/* and... the queue elements we'll be using */
static struct xmitQel qels[16];
/* and their corresponding mailboxes */
static unsigned char mailbox[16];
static unsigned char mboxinuse[16] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
static int wait_timeout(struct net_device *dev, int c)
{
/* returns true if it stayed c */
/* this uses base+6, but it's ok */
int i;
/* twenty second or so total */
for(i=0;i<200000;i++) {
if ( c != inb_p(dev->base_addr+6) ) return 0;
udelay(100);
}
return 1; /* timed out */
}
/* get the first free mailbox */
static int getmbox(void)
{
unsigned long flags;
int i;
spin_lock_irqsave(&mbox_lock, flags);
for(i=1;i<16;i++) if(!mboxinuse[i]) {
mboxinuse[i]=1;
spin_unlock_irqrestore(&mbox_lock, flags);
return i;
}
spin_unlock_irqrestore(&mbox_lock, flags);
return 0;
}
/* read a command from the card */
static void handlefc(struct net_device *dev)
{
/* called *only* from idle, non-reentrant */
int dma = dev->dma;
int base = dev->base_addr;
unsigned long flags;
flags=claim_dma_lock();
disable_dma(dma);
clear_dma_ff(dma);
set_dma_mode(dma,DMA_MODE_READ);
set_dma_addr(dma,virt_to_bus(ltdmacbuf));
set_dma_count(dma,50);
enable_dma(dma);
release_dma_lock(flags);
inb_p(base+3);
inb_p(base+2);
if ( wait_timeout(dev,0xfc) ) printk("timed out in handlefc\n");
}
/* read data from the card */
static void handlefd(struct net_device *dev)
{
int dma = dev->dma;
int base = dev->base_addr;
unsigned long flags;
flags=claim_dma_lock();
disable_dma(dma);
clear_dma_ff(dma);
set_dma_mode(dma,DMA_MODE_READ);
set_dma_addr(dma,virt_to_bus(ltdmabuf));
set_dma_count(dma,800);
enable_dma(dma);
release_dma_lock(flags);
inb_p(base+3);
inb_p(base+2);
if ( wait_timeout(dev,0xfd) ) printk("timed out in handlefd\n");
sendup_buffer(dev);
}
static void handlewrite(struct net_device *dev)
{
/* called *only* from idle, non-reentrant */
/* on entry, 0xfb and ltdmabuf holds data */
int dma = dev->dma;
int base = dev->base_addr;
unsigned long flags;
flags=claim_dma_lock();
disable_dma(dma);
clear_dma_ff(dma);
set_dma_mode(dma,DMA_MODE_WRITE);
set_dma_addr(dma,virt_to_bus(ltdmabuf));
set_dma_count(dma,800);
enable_dma(dma);
release_dma_lock(flags);
inb_p(base+3);
inb_p(base+2);
if ( wait_timeout(dev,0xfb) ) {
flags=claim_dma_lock();
printk("timed out in handlewrite, dma res %d\n",
get_dma_residue(dev->dma) );
release_dma_lock(flags);
}
}
static void handleread(struct net_device *dev)
{
/* on entry, 0xfb */
/* on exit, ltdmabuf holds data */
int dma = dev->dma;
int base = dev->base_addr;
unsigned long flags;
flags=claim_dma_lock();
disable_dma(dma);
clear_dma_ff(dma);
set_dma_mode(dma,DMA_MODE_READ);
set_dma_addr(dma,virt_to_bus(ltdmabuf));
set_dma_count(dma,800);
enable_dma(dma);
release_dma_lock(flags);
inb_p(base+3);
inb_p(base+2);
if ( wait_timeout(dev,0xfb) ) printk("timed out in handleread\n");
}
static void handlecommand(struct net_device *dev)
{
/* on entry, 0xfa and ltdmacbuf holds command */
int dma = dev->dma;
int base = dev->base_addr;
unsigned long flags;
flags=claim_dma_lock();
disable_dma(dma);
clear_dma_ff(dma);
set_dma_mode(dma,DMA_MODE_WRITE);
set_dma_addr(dma,virt_to_bus(ltdmacbuf));
set_dma_count(dma,50);
enable_dma(dma);
release_dma_lock(flags);
inb_p(base+3);
inb_p(base+2);
if ( wait_timeout(dev,0xfa) ) printk("timed out in handlecommand\n");
}
/* ready made command for getting the result from the card */
static unsigned char rescbuf[2] = {LT_GETRESULT,0};
static unsigned char resdbuf[2];
static int QInIdle;
/* idle expects to be called with the IRQ line high -- either because of
* an interrupt, or because the line is tri-stated
*/
static void idle(struct net_device *dev)
{
unsigned long flags;
int state;
/* FIXME This is initialized to shut the warning up, but I need to
* think this through again.
*/
struct xmitQel *q = NULL;
int oops;
int i;
int base = dev->base_addr;
spin_lock_irqsave(&txqueue_lock, flags);
if(QInIdle) {
spin_unlock_irqrestore(&txqueue_lock, flags);
return;
}
QInIdle = 1;
spin_unlock_irqrestore(&txqueue_lock, flags);
/* this tri-states the IRQ line */
(void) inb_p(base+6);
oops = 100;
loop:
if (0>oops--) {
printk("idle: looped too many times\n");
goto done;
}
state = inb_p(base+6);
if (state != inb_p(base+6)) goto loop;
switch(state) {
case 0xfc:
/* incoming command */
if (debug & DEBUG_LOWER) printk("idle: fc\n");
handlefc(dev);
break;
case 0xfd:
/* incoming data */
if(debug & DEBUG_LOWER) printk("idle: fd\n");
handlefd(dev);
break;
case 0xf9:
/* result ready */
if (debug & DEBUG_LOWER) printk("idle: f9\n");
if(!mboxinuse[0]) {
mboxinuse[0] = 1;
qels[0].cbuf = rescbuf;
qels[0].cbuflen = 2;
qels[0].dbuf = resdbuf;
qels[0].dbuflen = 2;
qels[0].QWrite = 0;
qels[0].mailbox = 0;
enQ(&qels[0]);
}
inb_p(dev->base_addr+1);
inb_p(dev->base_addr+0);
if( wait_timeout(dev,0xf9) )
printk("timed out idle f9\n");
break;
case 0xf8:
/* ?? */
if (xmQhd) {
inb_p(dev->base_addr+1);
inb_p(dev->base_addr+0);
if(wait_timeout(dev,0xf8) )
printk("timed out idle f8\n");
} else {
goto done;
}
break;
case 0xfa:
/* waiting for command */
if(debug & DEBUG_LOWER) printk("idle: fa\n");
if (xmQhd) {
q=deQ();
memcpy(ltdmacbuf,q->cbuf,q->cbuflen);
ltdmacbuf[1] = q->mailbox;
if (debug>1) {
int n;
printk("ltpc: sent command ");
n = q->cbuflen;
if (n>100) n=100;
for(i=0;i<n;i++)
printk("%02x ",ltdmacbuf[i]);
printk("\n");
}
handlecommand(dev);
if(0xfa==inb_p(base+6)) {
/* we timed out, so return */
goto done;
}
} else {
/* we don't seem to have a command */
if (!mboxinuse[0]) {
mboxinuse[0] = 1;
qels[0].cbuf = rescbuf;
qels[0].cbuflen = 2;
qels[0].dbuf = resdbuf;
qels[0].dbuflen = 2;
qels[0].QWrite = 0;
qels[0].mailbox = 0;
enQ(&qels[0]);
} else {
printk("trouble: response command already queued\n");
goto done;
}
}
break;
case 0Xfb:
/* data transfer ready */
if(debug & DEBUG_LOWER) printk("idle: fb\n");
if(q->QWrite) {
memcpy(ltdmabuf,q->dbuf,q->dbuflen);
handlewrite(dev);
} else {
handleread(dev);
/* non-zero mailbox numbers are for
commmands, 0 is for GETRESULT
requests */
if(q->mailbox) {
memcpy(q->dbuf,ltdmabuf,q->dbuflen);
} else {
/* this was a result */
mailbox[ 0x0f & ltdmabuf[0] ] = ltdmabuf[1];
mboxinuse[0]=0;
}
}
break;
}
goto loop;
done:
QInIdle=0;
/* now set the interrupts back as appropriate */
/* the first read takes it out of tri-state (but still high) */
/* the second resets it */
/* note that after this point, any read of base+6 will
trigger an interrupt */
if (dev->irq) {
inb_p(base+7);
inb_p(base+7);
}
}
static int do_write(struct net_device *dev, void *cbuf, int cbuflen,
void *dbuf, int dbuflen)
{
int i = getmbox();
int ret;
if(i) {
qels[i].cbuf = cbuf;
qels[i].cbuflen = cbuflen;
qels[i].dbuf = dbuf;
qels[i].dbuflen = dbuflen;
qels[i].QWrite = 1;
qels[i].mailbox = i; /* this should be initted rather */
enQ(&qels[i]);
idle(dev);
ret = mailbox[i];
mboxinuse[i]=0;
return ret;
}
printk("ltpc: could not allocate mbox\n");
return -1;
}
static int do_read(struct net_device *dev, void *cbuf, int cbuflen,
void *dbuf, int dbuflen)
{
int i = getmbox();
int ret;
if(i) {
qels[i].cbuf = cbuf;
qels[i].cbuflen = cbuflen;
qels[i].dbuf = dbuf;
qels[i].dbuflen = dbuflen;
qels[i].QWrite = 0;
qels[i].mailbox = i; /* this should be initted rather */
enQ(&qels[i]);
idle(dev);
ret = mailbox[i];
mboxinuse[i]=0;
return ret;
}
printk("ltpc: could not allocate mbox\n");
return -1;
}
/* end of idle handlers -- what should be seen is do_read, do_write */
static struct timer_list ltpc_timer;
static netdev_tx_t ltpc_xmit(struct sk_buff *skb, struct net_device *dev);
static int read_30 ( struct net_device *dev)
{
lt_command c;
c.getflags.command = LT_GETFLAGS;
return do_read(dev, &c, sizeof(c.getflags),&c,0);
}
static int set_30 (struct net_device *dev,int x)
{
lt_command c;
c.setflags.command = LT_SETFLAGS;
c.setflags.flags = x;
return do_write(dev, &c, sizeof(c.setflags),&c,0);
}
/* LLAP to DDP translation */
static int sendup_buffer (struct net_device *dev)
{
/* on entry, command is in ltdmacbuf, data in ltdmabuf */
/* called from idle, non-reentrant */
int dnode, snode, llaptype, len;
int sklen;
struct sk_buff *skb;
struct lt_rcvlap *ltc = (struct lt_rcvlap *) ltdmacbuf;
if (ltc->command != LT_RCVLAP) {
printk("unknown command 0x%02x from ltpc card\n",ltc->command);
return -1;
}
dnode = ltc->dnode;
snode = ltc->snode;
llaptype = ltc->laptype;
len = ltc->length;
sklen = len;
if (llaptype == 1)
sklen += 8; /* correct for short ddp */
if(sklen > 800) {
printk(KERN_INFO "%s: nonsense length in ltpc command 0x14: 0x%08x\n",
dev->name,sklen);
return -1;
}
if ( (llaptype==0) || (llaptype>2) ) {
printk(KERN_INFO "%s: unknown LLAP type: %d\n",dev->name,llaptype);
return -1;
}
skb = dev_alloc_skb(3+sklen);
if (skb == NULL)
{
printk("%s: dropping packet due to memory squeeze.\n",
dev->name);
return -1;
}
skb->dev = dev;
if (sklen > len)
skb_reserve(skb,8);
skb_put(skb,len+3);
skb->protocol = htons(ETH_P_LOCALTALK);
/* add LLAP header */
skb->data[0] = dnode;
skb->data[1] = snode;
skb->data[2] = llaptype;
skb_reset_mac_header(skb); /* save pointer to llap header */
skb_pull(skb,3);
/* copy ddp(s,e)hdr + contents */
skb_copy_to_linear_data(skb, ltdmabuf, len);
skb_reset_transport_header(skb);
dev->stats.rx_packets++;
dev->stats.rx_bytes += skb->len;
/* toss it onwards */
netif_rx(skb);
return 0;
}
/* the handler for the board interrupt */
static irqreturn_t
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 21:55:46 +08:00
ltpc_interrupt(int irq, void *dev_id)
{
struct net_device *dev = dev_id;
if (dev==NULL) {
printk("ltpc_interrupt: unknown device.\n");
return IRQ_NONE;
}
inb_p(dev->base_addr+6); /* disable further interrupts from board */
idle(dev); /* handle whatever is coming in */
/* idle re-enables interrupts from board */
return IRQ_HANDLED;
}
/***
*
* The ioctls that the driver responds to are:
*
* SIOCSIFADDR -- do probe using the passed node hint.
* SIOCGIFADDR -- return net, node.
*
* some of this stuff should be done elsewhere.
*
***/
static int ltpc_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
{
struct sockaddr_at *sa = (struct sockaddr_at *) &ifr->ifr_addr;
/* we'll keep the localtalk node address in dev->pa_addr */
struct ltpc_private *ltpc_priv = netdev_priv(dev);
struct atalk_addr *aa = &ltpc_priv->my_addr;
struct lt_init c;
int ltflags;
if(debug & DEBUG_VERBOSE) printk("ltpc_ioctl called\n");
switch(cmd) {
case SIOCSIFADDR:
aa->s_net = sa->sat_addr.s_net;
/* this does the probe and returns the node addr */
c.command = LT_INIT;
c.hint = sa->sat_addr.s_node;
aa->s_node = do_read(dev,&c,sizeof(c),&c,0);
/* get all llap frames raw */
ltflags = read_30(dev);
ltflags |= LT_FLAG_ALLLAP;
set_30 (dev,ltflags);
dev->broadcast[0] = 0xFF;
dev->dev_addr[0] = aa->s_node;
dev->addr_len=1;
return 0;
case SIOCGIFADDR:
sa->sat_addr.s_net = aa->s_net;
sa->sat_addr.s_node = aa->s_node;
return 0;
default:
return -EINVAL;
}
}
static void set_multicast_list(struct net_device *dev)
{
/* This needs to be present to keep netatalk happy. */
/* Actually netatalk needs fixing! */
}
static int ltpc_poll_counter;
static void ltpc_poll(unsigned long l)
{
struct net_device *dev = (struct net_device *) l;
del_timer(&ltpc_timer);
if(debug & DEBUG_VERBOSE) {
if (!ltpc_poll_counter) {
ltpc_poll_counter = 50;
printk("ltpc poll is alive\n");
}
ltpc_poll_counter--;
}
if (!dev)
return; /* we've been downed */
/* poll 20 times per second */
idle(dev);
ltpc_timer.expires = jiffies + HZ/20;
add_timer(&ltpc_timer);
}
/* DDP to LLAP translation */
static netdev_tx_t ltpc_xmit(struct sk_buff *skb, struct net_device *dev)
{
/* in kernel 1.3.xx, on entry skb->data points to ddp header,
* and skb->len is the length of the ddp data + ddp header
*/
int i;
struct lt_sendlap cbuf;
unsigned char *hdr;
cbuf.command = LT_SENDLAP;
cbuf.dnode = skb->data[0];
cbuf.laptype = skb->data[2];
skb_pull(skb,3); /* skip past LLAP header */
cbuf.length = skb->len; /* this is host order */
skb_reset_transport_header(skb);
if(debug & DEBUG_UPPER) {
printk("command ");
for(i=0;i<6;i++)
printk("%02x ",((unsigned char *)&cbuf)[i]);
printk("\n");
}
hdr = skb_transport_header(skb);
do_write(dev, &cbuf, sizeof(cbuf), hdr, skb->len);
if(debug & DEBUG_UPPER) {
printk("sent %d ddp bytes\n",skb->len);
for (i = 0; i < skb->len; i++)
printk("%02x ", hdr[i]);
printk("\n");
}
dev->stats.tx_packets++;
dev->stats.tx_bytes += skb->len;
dev_kfree_skb(skb);
return NETDEV_TX_OK;
}
/* initialization stuff */
static int __init ltpc_probe_dma(int base, int dma)
{
int want = (dma == 3) ? 2 : (dma == 1) ? 1 : 3;
unsigned long timeout;
unsigned long f;
if (want & 1) {
if (request_dma(1,"ltpc")) {
want &= ~1;
} else {
f=claim_dma_lock();
disable_dma(1);
clear_dma_ff(1);
set_dma_mode(1,DMA_MODE_WRITE);
set_dma_addr(1,virt_to_bus(ltdmabuf));
set_dma_count(1,sizeof(struct lt_mem));
enable_dma(1);
release_dma_lock(f);
}
}
if (want & 2) {
if (request_dma(3,"ltpc")) {
want &= ~2;
} else {
f=claim_dma_lock();
disable_dma(3);
clear_dma_ff(3);
set_dma_mode(3,DMA_MODE_WRITE);
set_dma_addr(3,virt_to_bus(ltdmabuf));
set_dma_count(3,sizeof(struct lt_mem));
enable_dma(3);
release_dma_lock(f);
}
}
/* set up request */
/* FIXME -- do timings better! */
ltdmabuf[0] = LT_READMEM;
ltdmabuf[1] = 1; /* mailbox */
ltdmabuf[2] = 0; ltdmabuf[3] = 0; /* address */
ltdmabuf[4] = 0; ltdmabuf[5] = 1; /* read 0x0100 bytes */
ltdmabuf[6] = 0; /* dunno if this is necessary */
inb_p(io+1);
inb_p(io+0);
timeout = jiffies+100*HZ/100;
while(time_before(jiffies, timeout)) {
if ( 0xfa == inb_p(io+6) ) break;
}
inb_p(io+3);
inb_p(io+2);
while(time_before(jiffies, timeout)) {
if ( 0xfb == inb_p(io+6) ) break;
}
/* release the other dma channel (if we opened both of them) */
if ((want & 2) && (get_dma_residue(3)==sizeof(struct lt_mem))) {
want &= ~2;
free_dma(3);
}
if ((want & 1) && (get_dma_residue(1)==sizeof(struct lt_mem))) {
want &= ~1;
free_dma(1);
}
if (!want)
return 0;
return (want & 2) ? 3 : 1;
}
static const struct net_device_ops ltpc_netdev = {
.ndo_start_xmit = ltpc_xmit,
.ndo_do_ioctl = ltpc_ioctl,
.ndo_set_rx_mode = set_multicast_list,
};
struct net_device * __init ltpc_probe(void)
{
struct net_device *dev;
int err = -ENOMEM;
int x=0,y=0;
int autoirq;
unsigned long f;
unsigned long timeout;
dev = alloc_ltalkdev(sizeof(struct ltpc_private));
if (!dev)
goto out;
/* probe for the I/O port address */
if (io != 0x240 && request_region(0x220,8,"ltpc")) {
x = inb_p(0x220+6);
if ( (x!=0xff) && (x>=0xf0) ) {
io = 0x220;
goto got_port;
}
release_region(0x220,8);
}
if (io != 0x220 && request_region(0x240,8,"ltpc")) {
y = inb_p(0x240+6);
if ( (y!=0xff) && (y>=0xf0) ){
io = 0x240;
goto got_port;
}
release_region(0x240,8);
}
/* give up in despair */
printk(KERN_ERR "LocalTalk card not found; 220 = %02x, 240 = %02x.\n", x,y);
err = -ENODEV;
goto out1;
got_port:
/* probe for the IRQ line */
if (irq < 2) {
unsigned long irq_mask;
irq_mask = probe_irq_on();
/* reset the interrupt line */
inb_p(io+7);
inb_p(io+7);
/* trigger an interrupt (I hope) */
inb_p(io+6);
mdelay(2);
autoirq = probe_irq_off(irq_mask);
if (autoirq == 0) {
printk(KERN_ERR "ltpc: probe at %#x failed to detect IRQ line.\n", io);
} else {
irq = autoirq;
}
}
/* allocate a DMA buffer */
ltdmabuf = (unsigned char *) dma_mem_alloc(1000);
if (!ltdmabuf) {
printk(KERN_ERR "ltpc: mem alloc failed\n");
err = -ENOMEM;
goto out2;
}
ltdmacbuf = &ltdmabuf[800];
if(debug & DEBUG_VERBOSE) {
printk("ltdmabuf pointer %08lx\n",(unsigned long) ltdmabuf);
}
/* reset the card */
inb_p(io+1);
inb_p(io+3);
msleep(20);
inb_p(io+0);
inb_p(io+2);
inb_p(io+7); /* clear reset */
inb_p(io+4);
inb_p(io+5);
inb_p(io+5); /* enable dma */
inb_p(io+6); /* tri-state interrupt line */
ssleep(1);
/* now, figure out which dma channel we're using, unless it's
already been specified */
/* well, 0 is a legal DMA channel, but the LTPC card doesn't
use it... */
dma = ltpc_probe_dma(io, dma);
if (!dma) { /* no dma channel */
printk(KERN_ERR "No DMA channel found on ltpc card.\n");
err = -ENODEV;
goto out3;
}
/* print out friendly message */
if(irq)
printk(KERN_INFO "Apple/Farallon LocalTalk-PC card at %03x, IR%d, DMA%d.\n",io,irq,dma);
else
printk(KERN_INFO "Apple/Farallon LocalTalk-PC card at %03x, DMA%d. Using polled mode.\n",io,dma);
dev->netdev_ops = &ltpc_netdev;
dev->base_addr = io;
dev->irq = irq;
dev->dma = dma;
/* the card will want to send a result at this point */
/* (I think... leaving out this part makes the kernel crash,
so I put it back in...) */
f=claim_dma_lock();
disable_dma(dma);
clear_dma_ff(dma);
set_dma_mode(dma,DMA_MODE_READ);
set_dma_addr(dma,virt_to_bus(ltdmabuf));
set_dma_count(dma,0x100);
enable_dma(dma);
release_dma_lock(f);
(void) inb_p(io+3);
(void) inb_p(io+2);
timeout = jiffies+100*HZ/100;
while(time_before(jiffies, timeout)) {
if( 0xf9 == inb_p(io+6))
break;
schedule();
}
if(debug & DEBUG_VERBOSE) {
printk("setting up timer and irq\n");
}
/* grab it and don't let go :-) */
if (irq && request_irq( irq, ltpc_interrupt, 0, "ltpc", dev) >= 0)
{
(void) inb_p(io+7); /* enable interrupts from board */
(void) inb_p(io+7); /* and reset irq line */
} else {
if( irq )
printk(KERN_ERR "ltpc: IRQ already in use, using polled mode.\n");
dev->irq = 0;
/* polled mode -- 20 times per second */
/* this is really, really slow... should it poll more often? */
init_timer(&ltpc_timer);
ltpc_timer.function=ltpc_poll;
ltpc_timer.data = (unsigned long) dev;
ltpc_timer.expires = jiffies + HZ/20;
add_timer(&ltpc_timer);
}
err = register_netdev(dev);
if (err)
goto out4;
return NULL;
out4:
del_timer_sync(&ltpc_timer);
if (dev->irq)
free_irq(dev->irq, dev);
out3:
free_pages((unsigned long)ltdmabuf, get_order(1000));
out2:
release_region(io, 8);
out1:
free_netdev(dev);
out:
return ERR_PTR(err);
}
#ifndef MODULE
/* handles "ltpc=io,irq,dma" kernel command lines */
static int __init ltpc_setup(char *str)
{
int ints[5];
str = get_options(str, ARRAY_SIZE(ints), ints);
if (ints[0] == 0) {
if (str && !strncmp(str, "auto", 4)) {
/* do nothing :-) */
}
else {
/* usage message */
printk (KERN_ERR
"ltpc: usage: ltpc=auto|iobase[,irq[,dma]]\n");
return 0;
}
} else {
io = ints[1];
if (ints[0] > 1) {
irq = ints[2];
}
if (ints[0] > 2) {
dma = ints[3];
}
/* ignore any other parameters */
}
return 1;
}
__setup("ltpc=", ltpc_setup);
#endif /* MODULE */
static struct net_device *dev_ltpc;
#ifdef MODULE
MODULE_LICENSE("GPL");
module_param(debug, int, 0);
module_param(io, int, 0);
module_param(irq, int, 0);
module_param(dma, int, 0);
static int __init ltpc_module_init(void)
{
if(io == 0)
printk(KERN_NOTICE
"ltpc: Autoprobing is not recommended for modules\n");
dev_ltpc = ltpc_probe();
return PTR_RET(dev_ltpc);
}
module_init(ltpc_module_init);
#endif
static void __exit ltpc_cleanup(void)
{
if(debug & DEBUG_VERBOSE) printk("unregister_netdev\n");
unregister_netdev(dev_ltpc);
ltpc_timer.data = 0; /* signal the poll routine that we're done */
del_timer_sync(&ltpc_timer);
if(debug & DEBUG_VERBOSE) printk("freeing irq\n");
if (dev_ltpc->irq)
free_irq(dev_ltpc->irq, dev_ltpc);
if(debug & DEBUG_VERBOSE) printk("freeing dma\n");
if (dev_ltpc->dma)
free_dma(dev_ltpc->dma);
if(debug & DEBUG_VERBOSE) printk("freeing ioaddr\n");
if (dev_ltpc->base_addr)
release_region(dev_ltpc->base_addr,8);
free_netdev(dev_ltpc);
if(debug & DEBUG_VERBOSE) printk("free_pages\n");
free_pages( (unsigned long) ltdmabuf, get_order(1000));
if(debug & DEBUG_VERBOSE) printk("returning from cleanup_module\n");
}
module_exit(ltpc_cleanup);