2011-04-28 09:02:28

by Jan Beulich

[permalink] [raw]
Subject: [PATCH, v2] i386: tighten dependencies of CPU_SUP_*_32

Building in support for either of these CPUs is pointless when e.g.
M686 was selected (as such a kernel would use cmov instructions, which
aren't available on these older CPUs).

v2: Drop (inapprorpiate) change to CPU_SUP_TRANSMETA_32.

Signed-off-by: Jan Beulich <[email protected]>

---
arch/x86/Kconfig.cpu | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

--- 2.6.39-rc5/arch/x86/Kconfig.cpu
+++ 2.6.39-rc5-i386-cpu-sup-dependencies/arch/x86/Kconfig.cpu
@@ -440,7 +440,7 @@ config CPU_SUP_INTEL
config CPU_SUP_CYRIX_32
default y
bool "Support Cyrix processors" if PROCESSOR_SELECT
- depends on !64BIT
+ depends on M386 || M486 || M586 || M586TSC || M586MMX || (EXPERT && !64BIT)
---help---
This enables detection, tunings and quirks for Cyrix processors

@@ -494,7 +494,7 @@ config CPU_SUP_TRANSMETA_32
config CPU_SUP_UMC_32
default y
bool "Support UMC processors" if PROCESSOR_SELECT
- depends on !64BIT
+ depends on M386 || M486 || (EXPERT && !64BIT)
---help---
This enables detection, tunings and quirks for UMC processors




2011-04-28 16:42:57

by Dave Jones

[permalink] [raw]
Subject: Re: [PATCH, v2] i386: tighten dependencies of CPU_SUP_*_32

On Thu, Apr 28, 2011 at 10:03:39AM +0100, Jan Beulich wrote:

> --- 2.6.39-rc5/arch/x86/Kconfig.cpu
> +++ 2.6.39-rc5-i386-cpu-sup-dependencies/arch/x86/Kconfig.cpu
> @@ -440,7 +440,7 @@ config CPU_SUP_INTEL
> config CPU_SUP_CYRIX_32
> default y
> bool "Support Cyrix processors" if PROCESSOR_SELECT
> - depends on !64BIT
> + depends on M386 || M486 || M586 || M586TSC || M586MMX || (EXPERT && !64BIT)
> ---help---
> This enables detection, tunings and quirks for Cyrix processors
>
> @@ -494,7 +494,7 @@ config CPU_SUP_TRANSMETA_32
> config CPU_SUP_UMC_32
> default y
> bool "Support UMC processors" if PROCESSOR_SELECT
> - depends on !64BIT
> + depends on M386 || M486 || (EXPERT && !64BIT)
> ---help---
> This enables detection, tunings and quirks for UMC processors

PROCESSOR_SELECT is already dependant on EXPERT being set.

Dave

2011-04-29 06:53:23

by Jan Beulich

[permalink] [raw]
Subject: Re: [PATCH, v2] i386: tighten dependencies of CPU_SUP_*_32

>>> On 28.04.11 at 18:42, Dave Jones <[email protected]> wrote:
> On Thu, Apr 28, 2011 at 10:03:39AM +0100, Jan Beulich wrote:
>
> > --- 2.6.39-rc5/arch/x86/Kconfig.cpu
> > +++ 2.6.39-rc5-i386-cpu-sup-dependencies/arch/x86/Kconfig.cpu
> > @@ -440,7 +440,7 @@ config CPU_SUP_INTEL
> > config CPU_SUP_CYRIX_32
> > default y
> > bool "Support Cyrix processors" if PROCESSOR_SELECT
> > - depends on !64BIT
> > + depends on M386 || M486 || M586 || M586TSC || M586MMX || (EXPERT && !64BIT)
> > ---help---
> > This enables detection, tunings and quirks for Cyrix processors
> >
> > @@ -494,7 +494,7 @@ config CPU_SUP_TRANSMETA_32
> > config CPU_SUP_UMC_32
> > default y
> > bool "Support UMC processors" if PROCESSOR_SELECT
> > - depends on !64BIT
> > + depends on M386 || M486 || (EXPERT && !64BIT)
> > ---help---
> > This enables detection, tunings and quirks for UMC processors
>
> PROCESSOR_SELECT is already dependant on EXPERT being set.

If EXPERT was omitted from the depends condition, then the whole
option would be unavailable (and hence the prompt invisible) even
if PROCESSOR_SELECT=y (and e.g. M686=y).

The intention, however, is to allow the option to be selectable in
an EXPERT configuration (to have a temporary workaround in
case the condition set isn't precise, or to do testing on
hypothetical new CPUs).

Or to put it differently, the intended behavior change is precisely
for the PROCESSOR_SELECT=n case, so that no impossible CPUs
get auto-enabled.

Jan