It has been requested that we should let x86 behave like
other architectures so we at config time decide for 32 or 64 bit.
The challenge here is to mere the Kconfig files and I have done
an attempt to do so.See patch below.
The diffstat looks nice:
arch/x86/{Kconfig.i386 => Kconfig} | 1167 ++++++++++++++-------------
arch/x86/Kconfig.cpu | 79 ++-
arch/x86/Kconfig.i386 | 1546 ++++++------------------------------
arch/x86/Kconfig.x86_64 | 1051 +++++-------------------
scripts/kconfig/Makefile | 7 +-
5 files changed, 1116 insertions(+), 2734 deletions(-)But the merge was in no way trivial so a couple of trained eyes
on the result would be good.
I am heading for bed and I did not manage to unify the cpufreq
Kconfig files (but it looks trivial).
This needs to be done before we can apply this patch.I merged all config options with prompts to the common Kconfig file
and the non-prompt config options were kept in Kconfig.$(ARCH)
In some case I added a depends on X86_{32,64} just to document
the dependency since it was given implicit.In Kconfig.cpu I actually left out the 32/64 dependency since I
could not see it hamr since the relevant symbols were anyway not set.When we have merged the Kconfig files we need to fix the Makefile.
But that is trivial. We shall check X86_64 / X86_32 as replacement
for ARCH in arch/x86/Makefile and I think we are done.Comments?
I could split it up in smaller chunks - at least the Kconfig.cpu
thing could be merged independently. But I dunno if it is worth it.Sam
---
arch/x86/{Kconfig.i386 => Kconfig} | 1167 ++++++++++++++-------------
arch/x86/Kconfig.cpu | 79 ++-
arch/x86/Kconfig.i386 | 1546 ++++++------------------------------
arch/x86/Kconfig.x86_64 | 1051 +++++-------------------
scripts/kconfig/Makefile | 7 +-
5 files changed, 1116 insertions(+), 2734 deletions(-)
rename arch/x86/{Kconfig.i386 =>...
Just a heads up...
I have started to redo the patch below in smaller chunks and expect
to finish this tomorrow.
The cpufreq Kconfig file is done so no need for help here.I hope to make the patches small enough so we can be confident to
apply them despite that we are heading -rc2.
It would be nice to get "make ARCH=x86" working before next
kernel release - and with the Kconfig files merged this is trivial.
I hope the x86 maintainers agree - but that yours decision.Sam
-
I looked over your patch yesterday and it looked sane to me, but I would
definitely prefer if it was broken up for reviewability.-hpa
-
Some parens: X86_32 && (HIGHMEM4G || HIGHMEM64G)?
J
-
Yes - thanks.
This particular X86_32 dependency is for documentation only since
HIGHMEM4G and HIGHMEM64G is a X86_32 thing only.
But depite this is should be correct.
I will audit the other dependencies for the same bug.Sam
-
| Paul Jackson | Re: cpuset-remove-sched-domain-hooks-from-cpusets |
| Chuck Ebbert | Why do so many machines need "noapic"? |
| Bart Van Assche | Integration of SCST in the mainstream Linux kernel |
| Linus Torvalds | Linux 2.6.21 |
| Jonathan Gray | CVS: cvs.openbsd.org: src |
| Claudio Jeker | CVS: cvs.openbsd.org: src |
| Tobias Stoeckmann | CVS: cvs.openbsd.org: src |
| Otto Moerbeek | CVS: cvs.openbsd.org: src |
| Andrea Arcangeli | missing kvm smp tlb flush in invlpg |
| Zhang, Xiantao | RE: [PATCH] qemu-kvm: Flush icache after dma operations for ia64 |
| sudhir kumar | kvm-autotest: The automation plans? |
| Joerg Roedel | [PATCH 0/3] KVM SoftMMU fixes |
| Justin Mattock | Re: [19.666764] Disabling IRQ #23 |
| Oliver Neukum | Re: Autosuspend and interrupt-context output |
| David Vrabel | New subsystems: Ultra-Wideband radio, Wireless USB and WiMedia LLC Protocol |
| Felipe Balbi | g_file_storage and buggy scsi implementaions |
