Messages in this thread | | | Date | Mon, 5 Nov 2012 21:57:33 -0800 (PST) | From | David Rientjes <> | Subject | Re: [PATCH 103/104] mm: remove depends on CONFIG_EXPERIMENTAL |
| |
On Mon, 5 Nov 2012, Kees Cook wrote:
> >>> diff --git a/mm/Kconfig b/mm/Kconfig > >>> index a3f8ddd..679945e 100644 > >>> --- a/mm/Kconfig > >>> +++ b/mm/Kconfig > >>> @@ -1,6 +1,5 @@ > >>> config SELECT_MEMORY_MODEL > >>> def_bool y > >>> - depends on EXPERIMENTAL || ARCH_SELECT_MEMORY_MODEL > >>> > >>> choice > >>> prompt "Memory model" > >> > >> I thought you agreed to only drop EXPERIMENTAL here in > >> http://marc.info/?l=linux-kernel&m=135103415901094 and leave > >> ARCH_SELECT_MEMORY_MODEL, which you've orphaned with the above, for phase > >> two of your effort? > > > > Ah! Yes, thanks. I'll restore that. > > Wait, no. This is an "OR". ARCH_SELECT_MEMORY_MODEL has no affect on > SELECT_MEMORY_MODEL if EXPERIMENTAL is always considered on. My > proposal was to deal with ARCH_SELECT_MEMORY_MODEL separately. Did I > misunderstand something? >
We're rehashing the same discussion as before? I left the earlier thread with the understanding that this would become
depends on ARCH_SELECT_MEMORY_MODEL
and then fix ARCH_SELECT_MEMORY_MODEL when people complain for configurations that actually allow you to configure the memory model. It never should have been short-circuited by EXPERIMENTAL in the first place, but enabling it to be configurable for everybody and orphaning ARCH_SELECT_MEMORY_MODEL doesn't sound appropriate. I think we should do some due diligence in actually making ARCH_SELECT_MEMORY_MODEL work so people are presented with a config that will work on their machines.
(This is independent of the rest of the series, we can certainly remove EXPERIMENTAL regardless of this decision, I simply think we should be leaving ARCH_SELECT_MEMORY_MODEL to prevent users with CONFIG_EXPERIMENTAL=n from being presented with a new ability to change their memory model that actually doesn't work for them.)
| |