i8259.c in big endian

i8259.c in big endian

[Atsushi Nemoto]

arch/mips/kernel/i8259.c seems not working in big endian.

Here is a patch to fix this.

--- linux-sgi-cvs/arch/mips/kernel/i8259.c	Mon Sep 10 02:43:01 2001
+++ linux.new/arch/mips/kernel/i8259.c	Thu Nov  8 15:40:03 2001
@@ -70,8 +70,13 @@
 static unsigned int cached_irq_mask = 0xffff;
 
 #define __byte(x,y) 	(((unsigned char *)&(y))[x])
+#ifdef __BIG_ENDIAN
+#define cached_21	(__byte(1,cached_irq_mask))
+#define cached_A1	(__byte(0,cached_irq_mask))
+#else
 #define cached_21	(__byte(0,cached_irq_mask))
 #define cached_A1	(__byte(1,cached_irq_mask))
+#endif
 
 void disable_8259A_irq(unsigned int irq)
 {
---
Atsushi Nemoto

Re: i8259.c in big endian

[James Simmons]

Also I like to see away to pass in a base offset. I have a mips device
which has a i8259 chip but its io is offseted by 0xb0000000.

On Thu, 8 Nov 2001, Atsushi Nemoto wrote:

> arch/mips/kernel/i8259.c seems not working in big endian.
> 
> Here is a patch to fix this.
> 
> --- linux-sgi-cvs/arch/mips/kernel/i8259.c	Mon Sep 10 02:43:01 2001
> +++ linux.new/arch/mips/kernel/i8259.c	Thu Nov  8 15:40:03 2001
> @@ -70,8 +70,13 @@
>  static unsigned int cached_irq_mask = 0xffff;
>  
>  #define __byte(x,y) 	(((unsigned char *)&(y))[x])
> +#ifdef __BIG_ENDIAN
> +#define cached_21	(__byte(1,cached_irq_mask))
> +#define cached_A1	(__byte(0,cached_irq_mask))
> +#else
>  #define cached_21	(__byte(0,cached_irq_mask))
>  #define cached_A1	(__byte(1,cached_irq_mask))
> +#endif
>  
>  void disable_8259A_irq(unsigned int irq)
>  {
> ---
> Atsushi Nemoto
> 

Re: i8259.c in big endian

[Ralf Baechle]

On Thu, Nov 08, 2001 at 09:45:02AM -0800, James Simmons wrote:

> Also I like to see away to pass in a base offset. I have a mips device
> which has a i8259 chip but its io is offseted by 0xb0000000.

Then it's almost certainly an legacy ISA device with it's ports in ISA space,
so set mips_io_port_base to an apropriate value or does that not work for
you?

  Ralf

Re: i8259.c in big endian

[James Simmons]

> > Also I like to see away to pass in a base offset. I have a mips device
> > which has a i8259 chip but its io is offseted by 0xb0000000.
> 
> Then it's almost certainly an legacy ISA device with it's ports in ISA space,
> so set mips_io_port_base to an apropriate value or does that not work for
> you?

The mips_io_port_base is 0xa0000000. Whereas the i8259 chip is at
0xb0000000. The 0xa000000 value could be wrong. I will give it a try. 

Re: i8259.c in big endian

[Ralf Baechle]

On Thu, Nov 08, 2001 at 12:20:20PM -0800, James Simmons wrote:

> > > which has a i8259 chip but its io is offseted by 0xb0000000.
> > 
> > Then it's almost certainly an legacy ISA device with it's ports in ISA space,
> > so set mips_io_port_base to an apropriate value or does that not work for
> > you?
> 
> The mips_io_port_base is 0xa0000000. Whereas the i8259 chip is at
> 0xb0000000. The 0xa000000 value could be wrong. I will give it a try. 

As your board must have RAM at physical address zero 0xa0000000 is almost
certainly a wrong value.

  Ralf

Re: i8259.c in big endian

[James Simmons]

> > The mips_io_port_base is 0xa0000000. Whereas the i8259 chip is at
> > 0xb0000000. The 0xa000000 value could be wrong. I will give it a try. 
> 
> As your board must have RAM at physical address zero 0xa0000000 is almost
> certainly a wrong value.

Your right. The address of 0xb000000 is bogus. This is the value from the
old code. I will migrate the code over to the i8259.c stuff now. Thanks. 

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[James Simmons]

> > > The mips_io_port_base is 0xa0000000. Whereas the i8259 chip is at
> > > 0xb0000000. The 0xa000000 value could be wrong. I will give it a try. 
> > 
> > As your board must have RAM at physical address zero 0xa0000000 is almost
> > certainly a wrong value.
> 
> Your right. The address of 0xb000000 is bogus. This is the value from the
> old code. I will migrate the code over to the i8259.c stuff now. Thanks. 

Actually looking threw other mips branches now I see what the 0xb000000
is. It is the isa_port_base. 

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Jun Sun]

James Simmons wrote:
> 
> > > > The mips_io_port_base is 0xa0000000. Whereas the i8259 chip is at
> > > > 0xb0000000. The 0xa000000 value could be wrong. I will give it a try.
> > >
> > > As your board must have RAM at physical address zero 0xa0000000 is almost
> > > certainly a wrong value.
> >
> > Your right. The address of 0xb000000 is bogus. This is the value from the
> > old code. I will migrate the code over to the i8259.c stuff now. Thanks.
> 
> Actually looking threw other mips branches now I see what the 0xb000000
> is. It is the isa_port_base.
> 

You are probably referring to isa_slot_offset?

isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
send him a patch to get rid of it (as if he can't do it himself :-0) ?

Jun

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[James Simmons]

> You are probably referring to isa_slot_offset?

Yes. I meant isa_slot_offset.

> isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> send him a patch to get rid of it (as if he can't do it himself :-0) ?

Oh.

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Jim Paris]

> You are probably referring to isa_slot_offset?
> 
> isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> send him a patch to get rid of it (as if he can't do it himself :-0) ?

How should it be properly done?

-jim

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Jun Sun]

Jim Paris wrote:
> 
> > You are probably referring to isa_slot_offset?
> >
> > isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> > send him a patch to get rid of it (as if he can't do it himself :-0) ?
> 
> How should it be properly done?
> 

Use an ax!

Jun

P.S., I meant "just delete it wherever it appears."

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Ralf Baechle]

On Fri, Nov 09, 2001 at 10:30:45AM -0800, Jun Sun wrote:

> isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> send him a patch to get rid of it (as if he can't do it himself :-0) ?

Nope.  Somebody could fix isa_{read,write}[bwl] to use isa_slot_offset.
Right now all the ISA functions are broken.  So in case you're ISA drivers
seem to work that's the proof that they're broken *evil grin* :-)

  Ralf

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Jun Sun]

Ralf Baechle wrote:
> 
> On Fri, Nov 09, 2001 at 10:30:45AM -0800, Jun Sun wrote:
> 
> > isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> > send him a patch to get rid of it (as if he can't do it himself :-0) ?
> 
> Nope.  Somebody could fix isa_{read,write}[bwl] to use isa_slot_offset.
> Right now all the ISA functions are broken.  So in case you're ISA drivers
> seem to work that's the proof that they're broken *evil grin* :-)

I doubt if there is any MIPS machine using standard PC ISA bus that is *not*
on a PCI bus ...

Jun

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Florian Laws]

On Mon, Nov 12, 2001 at 10:31:54AM -0800, Jun Sun wrote:
> Ralf Baechle wrote:
> > 
> > On Fri, Nov 09, 2001 at 10:30:45AM -0800, Jun Sun wrote:
> > 
> > > isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> > > send him a patch to get rid of it (as if he can't do it himself :-0) ?
> > 
> > Nope.  Somebody could fix isa_{read,write}[bwl] to use isa_slot_offset.
> > Right now all the ISA functions are broken.  So in case you're ISA drivers
> > seem to work that's the proof that they're broken *evil grin* :-)
> 
> I doubt if there is any MIPS machine using standard PC ISA bus that is *not*
> on a PCI bus ...

FWIW, there is.
There are some old Siemens RM400 models (1989 vintage) with R3000 CPU and ISA 
bus.

Pretty unlikely that they'll be supportet any time, though... :-(

Florian

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[Ralf Baechle]

On Mon, Nov 12, 2001 at 10:31:54AM -0800, Jun Sun wrote:

> > > isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> > > send him a patch to get rid of it (as if he can't do it himself :-0) ?
> > 
> > Nope.  Somebody could fix isa_{read,write}[bwl] to use isa_slot_offset.
> > Right now all the ISA functions are broken.  So in case you're ISA drivers
> > seem to work that's the proof that they're broken *evil grin* :-)
> 
> I doubt if there is any MIPS machine using standard PC ISA bus that is *not*
> on a PCI bus ...

Many more than you'd ever want to support :-(  And yes, EISA and VLB bus
also.  I even heared the rumour about SBUS in MIPS machines.

  Ralf

Re: [Linux-mips-kernel]Re: i8259.c in big endian

[James Simmons]

> > isa_slot_offset is an obselete garbage.  Can someone do Ralf's a favor and
> > send him a patch to get rid of it (as if he can't do it himself :-0) ?
> 
> Nope.  Somebody could fix isa_{read,write}[bwl] to use isa_slot_offset.
> Right now all the ISA functions are broken.  So in case you're ISA drivers
> seem to work that's the proof that they're broken *evil grin* :-)

I see you fixed this. Thank you :-)  

Re: i8259.c in big endian

[Ralf Baechle]

On Thu, Nov 08, 2001 at 03:47:02PM +0900, Atsushi Nemoto wrote:
> Date: Thu, 08 Nov 2001 15:47:02 +0900 (JST)
> To: linux-mips@oss.sgi.com
> Cc: ralf@oss.sgi.com
> Subject: i8259.c in big endian
> From: Atsushi Nemoto <nemoto@toshiba-tops.co.jp>
> 
> arch/mips/kernel/i8259.c seems not working in big endian.

Below the fix I'm going to checkin.  Highly untested as I'm sitting in a
plase 800km northeast of New Zealand :-)

  Ralf