[parisc-linux] Re: [PATCH] missing exports

[parisc-linux] Re: [PATCH] missing exports

[Geert Uytterhoeven]

On Fri, 9 Jul 2004, Kars de Jong wrote:
> On Wed, 2004-05-26 at 20:55, Kars de Jong wrote:
> > On Wed, 2004-05-26 at 10:16, Geert Uytterhoeven wrote:
> > > On Tue, 25 May 2004, Kars de Jong wrote:
> > > > the input device layer HIL keyboard driver working, etc... :-(
> > >
> > > Aren't some PARISC people working on that as well?
> >
> > Hmm, yes, apparently there is a PARISC patch for 2.6 which has a port of
> > the HP300 HIL driver to the input device layer.
> >
> > They also have a far more extensive HIL driver which handles the mouse
> > as well, but I'm not sure if the HIL in the later PARISC models is the
> > same as the HIL in the HP300 series on a device driver level. I'd have
> > to study the driver a little better.
>
> I have ported the PARISC HIL stuff to the HP300. Basically only the SDC
> driver needed to be changed a bit, all the rest was generic already.
>
> My kernel crashes if loadkeys is used (which I don't understand yet) but
> otherwise it seems to work a bit. The keyboard and mouse are properly
> detected, but when I read /dev/input/mice I don't get any events.
>
> What should I do with these files? Just add them to our repository, and
> worry about merging with the PARISC people later?

What about posting the patches to the Linux/m68k and Linux/PA-RISC lists?

> Oh, all of these are already present in the m68k-2_4 branch, although
> some of them have moved. How does that work in CVS?

CVS doesn't do move. You have to delete the file in the old place, and add it
at the new place.

> The following files are new for the HIL stuff:
>
> drivers/input/keyboard/hil_kbd.c
> drivers/input/keyboard/hilkbd.c
> drivers/input/misc/hp_sdc_rtc.c
> drivers/input/mouse/hil_ptr.c
> drivers/input/serio/hil_mlc.c
> drivers/input/serio/hp_sdc.c
> drivers/input/serio/hp_sdc_mlc.c
> include/linux/hil.h
> include/linux/hil_mlc.h
> include/linux/hp_sdc.h

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: [PATCH] missing exports

[Matthew Wilcox]

On Sat, Jul 10, 2004 at 11:09:34AM +0200, Geert Uytterhoeven wrote:
> > worry about merging with the PARISC people later?
> 
> What about posting the patches to the Linux/m68k and Linux/PA-RISC lists?

That'd be great.  More shared code!
It's also good incentive for us to get this stuff merged upstream.
Helge, you said you'd take care of this... what's the current status?

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain

Re: [parisc-linux] Re: [PATCH] missing exports

[Helge Deller]

On Saturday 10 July 2004 15:05, Matthew Wilcox wrote:
> On Sat, Jul 10, 2004 at 11:09:34AM +0200, Geert Uytterhoeven wrote:
> > > worry about merging with the PARISC people later?
> > 
> > What about posting the patches to the Linux/m68k and Linux/PA-RISC lists?
> 
> That'd be great.  More shared code!
> It's also good incentive for us to get this stuff merged upstream.
> Helge, you said you'd take care of this... what's the current status?

Last time I tried the HIL stuff for 2.6 was still in bad shape.
It worked unreliable and all the locking due to the timing issues still increases the generic load of the machine.

Our code in kernel 2.4 has similiar problems, but this I think we can live with.

Anyway, I would be happy to see any patches and merge them upstream.

Helge
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: [PATCH] missing exports

[Kars de Jong]

On Sat, 2004-07-10 at 15:55, Helge Deller wrote:
> On Saturday 10 July 2004 15:05, Matthew Wilcox wrote:
> > On Sat, Jul 10, 2004 at 11:09:34AM +0200, Geert Uytterhoeven wrote:
> > > > worry about merging with the PARISC people later?
> > > 
> > > What about posting the patches to the Linux/m68k and Linux/PA-RISC lists?
> > 
> > That'd be great.  More shared code!
> > It's also good incentive for us to get this stuff merged upstream.
> > Helge, you said you'd take care of this... what's the current status?
> 
> Last time I tried the HIL stuff for 2.6 was still in bad shape.
> It worked unreliable and all the locking due to the timing issues still increases the generic load of the machine.

Yes, I noticed that...

It seems to be worse than 2.4 though. On my HP9000/425t the keyboard
sometimes locks up for a while and then starts reacting again. The mouse
gets detected but doesn't seem to generate any events on
/dev/input/mice.

> Our code in kernel 2.4 has similiar problems, but this I think we can live with.
> 
> Anyway, I would be happy to see any patches and merge them upstream.

OK, here they are. Nothing to fix the reliability issues though, I'm
affraid.

I needed remove the __exit marker from hp_sdc_exit() because the linker
didn't like it being called from hp_sdc_register():

drivers/built-in.o: In function `hp_sdc_register':
drivers/built-in.o(.init.text+0x396e): undefined reference to `local symbols in discarded section .exit.text'

I don't really know what that means, seems like it doesn't like calling
an __exit function from an __init function.

ChangeLog:

drivers/input/keyboard/hil_kbd.c:
	Check for KEY_RESERVED _after_ translation.

include/linux/hp_sdc.h
drivers/input/serio/hp_sdc.c
	Add support for HP9000/[34]00 series.


Kind regards,

Kars.

diff -ur /tmp/parisc/drivers/input/keyboard/hil_kbd.c /usr/src/linux-2.6-m68k-hp/drivers/input/keyboard/hil_kbd.c
--- /tmp/parisc/drivers/input/keyboard/hil_kbd.c	Sat Jul 10 15:47:52 2004
+++ /usr/src/linux-2.6-m68k-hp/drivers/input/keyboard/hil_kbd.c	Sun Jun 27 13:26:21 2004
@@ -156,9 +156,9 @@
 			key = kbd->data[cnt++];
 			up = key & HIL_KBD_SET1_UPBIT;
 			key &= (~HIL_KBD_SET1_UPBIT & 0xff);
-			key = key >> HIL_KBD_SET1_SHIFT;
+			key = hil_kbd_set1[key >> HIL_KBD_SET1_SHIFT];
 			if (key != KEY_RESERVED)
-				input_report_key(dev, hil_kbd_set1[key], !up);
+				input_report_key(dev, key, !up);
 		}
 		break;
 	case HIL_POL_CHARTYPE_SET2:
@@ -180,9 +180,9 @@
 			key = kbd->data[cnt++];
 			up = key & HIL_KBD_SET3_UPBIT;
 			key &= (~HIL_KBD_SET1_UPBIT & 0xff);
-			key = key >> HIL_KBD_SET3_SHIFT;
+			key = hil_kbd_set3[key >> HIL_KBD_SET3_SHIFT];
 			if (key != KEY_RESERVED)
-				input_report_key(dev, hil_kbd_set3[key], !up);
+				input_report_key(dev, key, !up);
 		}
 		break;
 	}
diff -ur /tmp/parisc/drivers/input/serio/hp_sdc.c /usr/src/linux-2.6-m68k-hp/drivers/input/serio/hp_sdc.c
--- /tmp/parisc/drivers/input/serio/hp_sdc.c	Sat Jul 10 15:47:56 2004
+++ /usr/src/linux-2.6-m68k-hp/drivers/input/serio/hp_sdc.c	Sat Jul 10 18:22:19 2004
@@ -72,7 +72,20 @@
 #include <linux/hil.h>
 #include <asm/io.h>
 #include <asm/system.h>
-#include <asm/parisc-device.h>
+
+/* Machine-specific abstraction */
+
+#if defined(__hppa__)
+# include <asm/parisc-device.h>
+# define sdc_readb(p)		gsc_readb(p)
+# define sdc_writeb(v,p)	gsc_writeb((v),(p))
+#elif defined(__mc68000__)
+# include <asm/uaccess.h>
+# define sdc_readb(p)		in_8(p)
+# define sdc_writeb(v,p)	out_8((p),(v))
+#else
+# error "HIL is not supported on this platform"
+#endif
 
 #define PREFIX "HP SDC: "
 
@@ -99,7 +112,7 @@
 	unsigned long flags;
 
 	write_lock_irqsave(&hp_sdc.ibf_lock, flags);
-	status = gsc_readb(hp_sdc.status_io);
+	status = sdc_readb(hp_sdc.status_io);
 	if (!(status & HP_SDC_STATUS_IBF)) hp_sdc.ibf = 0;
 	write_unlock_irqrestore(&hp_sdc.ibf_lock, flags);
 
@@ -107,7 +120,7 @@
 }
 
 static inline uint8_t hp_sdc_data_in8 (void) {
-	return gsc_readb(hp_sdc.data_io); 
+	return sdc_readb(hp_sdc.data_io); 
 }
 
 static inline void hp_sdc_status_out8 (uint8_t val) {
@@ -116,7 +129,7 @@
 	write_lock_irqsave(&hp_sdc.ibf_lock, flags);
 	hp_sdc.ibf = 1;
 	if ((val & 0xf0) == 0xe0) hp_sdc.wi = 0xff;
-	gsc_writeb(val, hp_sdc.status_io);
+	sdc_writeb(val, hp_sdc.status_io);
 	write_unlock_irqrestore(&hp_sdc.ibf_lock, flags);
 }
 
@@ -125,7 +138,7 @@
 
 	write_lock_irqsave(&hp_sdc.ibf_lock, flags);
 	hp_sdc.ibf = 1;
-	gsc_writeb(val, hp_sdc.data_io);
+	sdc_writeb(val, hp_sdc.data_io);
 	write_unlock_irqrestore(&hp_sdc.ibf_lock, flags);
 }
 
@@ -146,7 +159,7 @@
 	}
 	read_unlock(lock);
 	write_lock(lock);
-	while (gsc_readb(hp_sdc.status_io) & HP_SDC_STATUS_IBF) {};
+	while (sdc_readb(hp_sdc.status_io) & HP_SDC_STATUS_IBF) {};
 	hp_sdc.ibf = 0;
 	write_unlock_irqrestore(lock, flags);
 }
@@ -636,6 +649,7 @@
 
 	hp_sdc.hil = callback;
 	hp_sdc.im &= ~HP_SDC_IM_HIL;
+	hp_sdc.im &= ~HP_SDC_IM_RESET;
 	hp_sdc.set_im = 1;
 	write_unlock_irq(&hp_sdc.hook_lock);
 
@@ -661,6 +675,7 @@
 	/* Enable interrupts from the HIL MLC */
 	hp_sdc.cooked = callback;
 	hp_sdc.im &= ~HP_SDC_IM_HIL;
+	hp_sdc.im &= ~HP_SDC_IM_RESET;
 	hp_sdc.set_im = 1;
 	write_unlock_irq(&hp_sdc.hook_lock);
 
@@ -705,6 +720,7 @@
 	/* Disable interrupts from HIL only if there is no cooked driver. */
 	if(hp_sdc.cooked == NULL) {
 		hp_sdc.im |= HP_SDC_IM_HIL;
+		hp_sdc.im |= HP_SDC_IM_RESET;
 		hp_sdc.set_im = 1;
 	}
 	write_unlock_irq(&hp_sdc.hook_lock);
@@ -727,6 +743,7 @@
 	/* Disable interrupts from HIL only if there is no raw HIL driver. */
 	if(hp_sdc.hil == NULL) {
 		hp_sdc.im |= HP_SDC_IM_HIL;
+		hp_sdc.im |= HP_SDC_IM_RESET;
 		hp_sdc.set_im = 1;
 	}
 	write_unlock_irq(&hp_sdc.hook_lock);
@@ -746,6 +763,8 @@
 
 /************************** Module Initialization ***************************/
 
+#if defined(__hppa__)
+
 static struct parisc_device_id hp_sdc_tbl[] = {
 	{
 		.hw_type =	HPHW_FIO, 
@@ -758,15 +777,17 @@
 
 MODULE_DEVICE_TABLE(parisc, hp_sdc_tbl);
 
-static int __init hp_sdc_init(struct parisc_device *d);
+static int __init hp_sdc_init_hppa(struct parisc_device *d);
 
 static struct parisc_driver hp_sdc_driver = {
 	.name =		"HP SDC",
 	.id_table =	hp_sdc_tbl,
-	.probe =	hp_sdc_init,
+	.probe =	hp_sdc_init_hppa,
 };
 
-static int __init hp_sdc_init(struct parisc_device *d)
+#endif /* __hppa__ */
+
+static int __init hp_sdc_init(void)
 {
 	int i;
 	char *errstr = NULL;
@@ -774,20 +795,8 @@
 	uint8_t ts_sync[6];
 	struct semaphore s_sync;
 
-
 	errstr = "foo\n";
 
-	if (!d) return 1;
-	if (hp_sdc.dev != NULL) return 1;	/* We only expect one SDC */
-
-	hp_sdc.dev = d;
-	hp_sdc.irq		= d->irq;
-	/* TODO: Is NMI == IRQ - 1 all cases, or is there a way to query? */
-	hp_sdc.nmi		= d->irq - 1;
-	hp_sdc.base_io		= (unsigned long) d->hpa;
-	hp_sdc.data_io		= (unsigned long) d->hpa + 0x800;
-	hp_sdc.status_io	= (unsigned long) d->hpa + 0x801;
-
   	hp_sdc.lock		= RW_LOCK_UNLOCKED;
   	hp_sdc.ibf_lock		= RW_LOCK_UNLOCKED;
   	hp_sdc.rtq_lock		= RW_LOCK_UNLOCKED;
@@ -822,6 +831,7 @@
 	hp_sdc.dev_err = -EBUSY;
 
 	errstr = "IO not available for";
+#if defined(__hppa__)
         if (request_region(hp_sdc.data_io, 2, hp_sdc_driver.name)) goto err0;
 
 	errstr = "IRQ not available for";
@@ -829,8 +839,19 @@
 		       (void *) hp_sdc.base_io)) goto err1;
 
 	errstr = "NMI not available for";
+
         if (request_irq(hp_sdc.nmi, &hp_sdc_nmisr, 0, "HP SDC NMI", 
 			(void*)d->hpa)) goto err2;
+#elif defined(__mc68000__)
+	errstr = "IRQ not available for";
+        if(request_irq(hp_sdc.irq, &hp_sdc_isr, 0, "HP SDC", 
+		       (void *) hp_sdc.base_io)) goto err1;
+
+	errstr = "NMI not available for";
+	
+	if (request_irq(hp_sdc.nmi, &hp_sdc_nmisr, 0, "HP SDC NMI", 
+			(void *) hp_sdc.base_io)) goto err2;
+#endif	
 
 	printk(KERN_INFO PREFIX "HP SDC at 0x%p, IRQ %d (NMI IRQ %d)\n", 
 	       (void *)hp_sdc.base_io, hp_sdc.irq, hp_sdc.nmi);
@@ -872,13 +893,37 @@
 	return hp_sdc.dev_err;
 }
 
+#if defined(__hppa__)
+
+static int __init hp_sdc_init_hppa(struct parisc_device *d)
+{
+	if (!d) return 1;
+	if (hp_sdc.dev != NULL) return 1;	/* We only expect one SDC */
+
+	hp_sdc.dev = d;
+	hp_sdc.irq		= d->irq;
+	/* TODO: Is NMI == IRQ - 1 all cases, or is there a way to query? */
+	hp_sdc.nmi		= d->irq - 1;
+	hp_sdc.base_io		= (unsigned long) d->hpa;
+	hp_sdc.data_io		= (unsigned long) d->hpa + 0x800;
+	hp_sdc.status_io	= (unsigned long) d->hpa + 0x801;
+
+	return hp_sdc_init();
+}
+
+#endif /* __hppa__ */
+
+#if !defined(__mc68000__) /* Link error on m68k! */
 static void __exit hp_sdc_exit(void)
+#else
+static void hp_sdc_exit(void)
+#endif
 {
 	write_lock_irq(&hp_sdc.lock);
 
 	/* Turn off all maskable "sub-function" irq's. */
 	hp_sdc_spin_ibf();
-	gsc_writeb(HP_SDC_CMD_SET_IM | HP_SDC_IM_MASK, hp_sdc.status_io);
+	sdc_writeb(HP_SDC_CMD_SET_IM | HP_SDC_IM_MASK, hp_sdc.status_io);
 
 	/* Wait until we know this has been processed by the i8042 */
 	hp_sdc_spin_ibf();
@@ -893,22 +938,45 @@
 
 /*        release_region(hp_sdc.data_io, 2); */
 
+#if defined(__hppa__)
 	if (unregister_parisc_driver(&hp_sdc_driver)) 
 		printk(KERN_WARNING PREFIX "Error unregistering HP SDC");
+#endif
 }
 
 static int __init hp_sdc_register(void)
 {
-	hp_sdc.dev = NULL;
-	hp_sdc.dev_err = 0;
 	hp_sdc_transaction tq_init;
 	uint8_t tq_init_seq[5];
 	struct semaphore tq_init_sem;
-
+#if defined(__mc68000__)
+	mm_segment_t fs;
+	unsigned char i;
+#endif
+	
+	hp_sdc.dev = NULL;
+	hp_sdc.dev_err = 0;
+#if defined(__hppa__)
 	if (register_parisc_driver(&hp_sdc_driver)) {
 		printk(KERN_WARNING PREFIX "Error registering SDC with system bus tree.\n");
 		return -ENODEV;
 	}
+#elif defined(__mc68000__)
+	if (!MACH_IS_HP300)
+	    return -ENODEV;
+
+	hp_sdc.irq	 = 1;
+	hp_sdc.nmi	 = 7;
+	hp_sdc.base_io	 = (unsigned long) 0xf0428000;
+	hp_sdc.data_io	 = (unsigned long) hp_sdc.base_io + 1;
+	hp_sdc.status_io = (unsigned long) hp_sdc.base_io + 3;
+	fs = get_fs();
+	set_fs(KERNEL_DS);
+	if (!get_user(i, (unsigned char *)hp_sdc.data_io))
+		hp_sdc.dev = (void *)1;
+	set_fs(fs);
+	hp_sdc.dev_err   = hp_sdc_init();
+#endif
 	if (hp_sdc.dev == NULL) {
 		printk(KERN_WARNING PREFIX "No SDC found.\n");
 		return hp_sdc.dev_err;
diff -ur /tmp/parisc/include/linux/hp_sdc.h /usr/src/linux-2.6-m68k-hp/include/linux/hp_sdc.h
--- /tmp/parisc/include/linux/hp_sdc.h	Sat Jul 10 15:47:56 2004
+++ /usr/src/linux-2.6-m68k-hp/include/linux/hp_sdc.h	Thu Jun 24 21:51:11 2004
@@ -42,7 +42,9 @@
 #include <linux/types.h>
 #include <linux/time.h>
 #include <linux/timer.h>
+#if defined(__hppa__)
 #include <asm/hardware.h>
+#endif
 
 
 /* No 4X status reads take longer than this (in usec).
@@ -284,6 +286,9 @@
 #ifdef __hppa__
 	struct parisc_device	*dev;
 	int		dev_err;	/* carries status from registration */
+#elif defined(__mc68000__)
+	void		*dev;
+	int		dev_err;	/* carries status from registration */	
 #else
 #error No support for device registration on this arch yet.
 #endif

Re: [parisc-linux] Re: [PATCH] missing exports

[Geert Uytterhoeven]

On Sat, 10 Jul 2004, Kars de Jong wrote:
> On Sat, 2004-07-10 at 15:55, Helge Deller wrote:
> > On Saturday 10 July 2004 15:05, Matthew Wilcox wrote:
> OK, here they are. Nothing to fix the reliability issues though, I'm
> affraid.
>
> I needed remove the __exit marker from hp_sdc_exit() because the linker
> didn't like it being called from hp_sdc_register():
>
> drivers/built-in.o: In function `hp_sdc_register':
> drivers/built-in.o(.init.text+0x396e): undefined reference to `local symbols in discarded section .exit.text'
>
> I don't really know what that means, seems like it doesn't like calling
> an __exit function from an __init function.

Indeed, __exit functions may be discarded.

> --- /tmp/parisc/include/linux/hp_sdc.h	Sat Jul 10 15:47:56 2004
> +++ /usr/src/linux-2.6-m68k-hp/include/linux/hp_sdc.h	Thu Jun 24 21:51:11 2004
> @@ -284,6 +286,9 @@
>  #ifdef __hppa__
>  	struct parisc_device	*dev;
>  	int		dev_err;	/* carries status from registration */
> +#elif defined(__mc68000__)
> +	void		*dev;
> +	int		dev_err;	/* carries status from registration */

You can merge dev_err as well.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: [parisc-linux] Re: [PATCH] missing exports

[Grant Grundler]

On Sat, Jul 10, 2004 at 06:36:30PM +0200, Kars de Jong wrote:
...
> @@ -284,6 +286,9 @@
>  #ifdef __hppa__
>  	struct parisc_device	*dev;
>  	int		dev_err;	/* carries status from registration */
> +#elif defined(__mc68000__)
> +	void		*dev;

Can *dev be unified eventually to "struct device"?

"void *" isn't my favorite construct.

grant

Re: [parisc-linux] Re: [PATCH] missing exports

[Helge Deller]

On Saturday 10 July 2004 18:36, Kars de Jong wrote:
> It seems to be worse than 2.4 though. On my HP9000/425t the keyboard
> sometimes locks up for a while and then starts reacting again. The mouse
> gets detected but doesn't seem to generate any events on
> /dev/input/mice.

Yes. Same problem here on HPPA/2.6 as well.

> > Anyway, I would be happy to see any patches and merge them upstream.
> 
> OK, here they are. Nothing to fix the reliability issues though, I'm
> affraid.

I applied your changes to the PARISC CVS tree, but as long as the mouse isn't working I think it's a bad idea to merge it upstream.

Helge
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: [PATCH] missing exports

[Matthew Wilcox]

On Mon, Jul 12, 2004 at 12:16:03AM +0200, Helge Deller wrote:
> I applied your changes to the PARISC CVS tree, but as long as the mouse isn't working I think it's a bad idea to merge it upstream.

I'm not sure that's true.  I think it's fine to merge it upstream and
fix it later.  Maybe someone else will notice the problem and fix it
for us ;-)

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain