I386 cli

I386 cli

[David Grothe]

In 2.5.41every architecture except Intel 386 has a "#define cli 
<something>" in its asm-arch/system.h file.  Is there supposed to be such a 
define in asm-i386/system.h?  If not, where does the "official" definition 
of cli() live for Intel?  Or what is the include file that one needs to 
pick it up?  I can't find it.
Thanks,
Dave

Re: I386 cli

[erich]

David Grothe <dave@gcom.com> wrote:

> In 2.5.41every architecture except Intel 386 has a "#define cli 
> <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> define in asm-i386/system.h?  If not, where does the "official" definition 
> of cli() live for Intel?  Or what is the include file that one needs to 
> pick it up?  I can't find it.

I'm sure there is no definition because "cli" is the native assembler
instruction on x86.

--
    Erich Stefan Boleyn     <erich@uruk.org>     http://www.uruk.org/
"Reality is truly stranger than fiction; Probably why fiction is so popular"

Re: I386 cli

[Vojtech Pavlik]

On Tue, Oct 22, 2002 at 03:01:47PM -0500, David Grothe wrote:
> In 2.5.41every architecture except Intel 386 has a "#define cli 
> <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> define in asm-i386/system.h?  If not, where does the "official" definition 
> of cli() live for Intel?  Or what is the include file that one needs to 
> pick it up?  I can't find it.

cli() is dead.

-- 
Vojtech Pavlik
SuSE Labs

Re: I386 cli

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 521 bytes --]

On Tue, 2002-10-22 at 22:01, David Grothe wrote:
> In 2.5.41every architecture except Intel 386 has a "#define cli 
> <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> define in asm-i386/system.h?  If not, where does the "official" definition 
> of cli() live for Intel?  Or what is the include file that one needs to 
> pick it up?  I can't find it.

cli no longer exists in the kernel
the other architectures have a broken definition probably that needs to
go away as well.




[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: I386 cli

[Sam Ravnborg]

On Tue, Oct 22, 2002 at 03:01:47PM -0500, David Grothe wrote:
> In 2.5.41every architecture except Intel 386 has a "#define cli 
> <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> define in asm-i386/system.h?  If not, where does the "official" definition 
> of cli() live for Intel?  Or what is the include file that one needs to 
> pick it up?  I can't find it.

I would advise you to read: Documentation/cli-sti-removal.txt
[I recall someone mentioned this documents were slightly out-dated - Ingo?]

The short version is that cli() is no loger valid, drivers using it does not compile.

	Sam

Re: I386 cli

[Alan Cox]

On Tue, 2002-10-22 at 21:01, David Grothe wrote:
> In 2.5.41every architecture except Intel 386 has a "#define cli 
> <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> define in asm-i386/system.h?  If not, where does the "official" definition 
> of cli() live for Intel?  Or what is the include file that one needs to 
> pick it up?  I can't find it.

The old style cli/global irq lock stuff has been exterminated. There is
no direct equivalent any more. 

Re: I386 cli

[Vojtech Pavlik]

On Tue, Oct 22, 2002 at 01:08:43PM -0700, erich@uruk.org wrote:

> David Grothe <dave@gcom.com> wrote:
> 
> > In 2.5.41every architecture except Intel 386 has a "#define cli 
> > <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> > define in asm-i386/system.h?  If not, where does the "official" definition 
> > of cli() live for Intel?  Or what is the include file that one needs to 
> > pick it up?  I can't find it.
> 
> I'm sure there is no definition because "cli" is the native assembler
> instruction on x86.

Wrong reason. Furthermore, cli(), meaning 'global interrupt disable,
across all processors', is not doable with a single instruction anyway.
It's not defined, because it should not be used - usually the usage of
cli() means a bug.

-- 
Vojtech Pavlik
SuSE Labs

Re: I386 cli

[erich]

erich@uruk.org wrote:

> David Grothe <dave@gcom.com> wrote:
> 
> > In 2.5.41every architecture except Intel 386 has a "#define cli 
> > <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> > define in asm-i386/system.h?  If not, where does the "official" definition 
> > of cli() live for Intel?  Or what is the include file that one needs to 
> > pick it up?  I can't find it.
> 
> I'm sure there is no definition because "cli" is the native assembler
> instruction on x86.

Whoops, foot-in-mouth...  those were C level definitions, not assembler
macros.

--
    Erich Stefan Boleyn     <erich@uruk.org>     http://www.uruk.org/
"Reality is truly stranger than fiction; Probably why fiction is so popular"

Re: I386 cli

[erich]

Vojtech Pavlik <vojtech@suse.cz> wrote:

...
> > I'm sure there is no definition because "cli" is the native assembler
> > instruction on x86.
> 
> Wrong reason. Furthermore, cli(), meaning 'global interrupt disable,
> across all processors', is not doable with a single instruction anyway.
> It's not defined, because it should not be used - usually the usage of
> cli() means a bug.

Yeah, I noticed I made a thinko here by not looking at the file itself,
and assuming it was just a direct assembler hack...  essentially the
C-level variant of what is called in most OSes "splhi/spllo" rather
than "cli/sti", probably because that was how it was originally
used.

Good that it's finally getting purged.

--
    Erich Stefan Boleyn     <erich@uruk.org>     http://www.uruk.org/
"Reality is truly stranger than fiction; Probably why fiction is so popular"

Re: I386 cli

[David Grothe]

Thanks, Sam.  That's the answer that I was looking for.  Wish I had thought 
to look in the Documentation directory first so as not to have to bother you.
-- Dave

At 10:19 PM 10/22/2002 Tuesday, Sam Ravnborg wrote:

>I would advise you to read: Documentation/cli-sti-removal.txt
>[I recall someone mentioned this documents were slightly out-dated - Ingo?]
>
>The short version is that cli() is no loger valid, drivers using it does 
>not compile.
>
>         Sam

Re: I386 cli

[Richard B. Johnson]

On Tue, 22 Oct 2002, David Grothe wrote:

> In 2.5.41every architecture except Intel 386 has a "#define cli 
> <something>" in its asm-arch/system.h file.  Is there supposed to be such a 
> define in asm-i386/system.h?  If not, where does the "official" definition 
> of cli() live for Intel?  Or what is the include file that one needs to 
> pick it up?  I can't find it.
> Thanks,
> Dave

Grep for 'spin_lock'.  You can't just use `cli` alone. It won't protect
the critical section. You probably need:

	spin_lock_irqsave(&lock, flags);
        //.... critical section
	spin_unlock_irqrestore(&lock, flags);


Cheers,
Dick Johnson
Penguin : Linux version 2.4.18 on an i686 machine (797.90 BogoMips).
   Bush : The Fourth Reich of America