Re: [PATCH] x86: introduce init_memory_mapping for 32bit

[Ingo Molnar <mingo@elte.hu>]

* Yinghai Lu <yhlu.kernel@gmail.com> wrote:

> so could use mem below max_low_pfn as early. could move several 
> function more early instead of waiting after paging_init including 
> moving relocate_initrd early, and kva related early done in 
> initmem_init

applied to tip/x86/setup-memory - thanks Yinghai.

a sidenote:

>  6 files changed, 128 insertions(+), 74 deletions(-)

this patch is too large - if it causes any problems it will not be very 
easy to figure out which exact change caused the problems.

Lets hope it goes all fine - but in the future lets try doing 
more+smaller patches, especially if they change some known-dangerous 
area of the kernel.

For example here a better splitup would have been to do 5 or more 
patches:

 1) first introduce init_memory_mapping() [but dont use it anywhere]
 2) add the init_memory_mapping() call to setup_arch()
 3) move remap_numa_kva()
 4) move relocate_initrd()
 5) remove the now unnecessary setup from paging_init()

... or something like that. The point is to manage risk: if there's 
multiple problem areas that a change is touching, try to isolate them 
from each other and introduce the change gradually.

The end result is still exactly the same, but much more 
reviewable/debuggable/bisectable.

	Ingo