Re: Linux 2.6.7-rc1

[Jeff Garzik <jgarzik@pobox.com>]

William Lee Irwin III wrote:
> William Lee Irwin III wrote:
> 
>>>I wouldn't qualify either of the major VM patch series merged as
>>>rewrites. I saw:
>>>(1) move unmapping function/helpers to different algorithm to save space
>>>(2) NUMA API and support functions
> 
> 
> On Sun, May 23, 2004 at 05:30:55PM -0400, Jeff Garzik wrote:
> 
>>You missed the pte chains going away, a fundamental change in the way 
>>reverse mapping is done?
> 
> 
> (1) describes that in more detail than "pte_chains going away". It's
> just a search algorithm. For anonymous pages, anon_vma just strobes
> offsets into vma inheritance chains, and anonmm just strobes vaddrs in
> all forks between execs, for file-backed memory they both use ->i_mmap.
> Yes, they can both be summarized in the same sentence. The scope of the
> changes are very limited and the presentation of them is a very clearly
> documented and incremental series of small changes.


You're getting lost in the details...  the scope of the anonvma change 
is _everybody_.  If you accept the premise that we are in a _stable_ 
kernel series, the core goal should be stability.  Non-trivial VM 
changes like this do not enhance stability in the short term, even if 
they are a good idea in the long run.  It's the whole idea behind 
minimizing changes and breakage in a stable series.

You've got all the major Linux vendors preparing (or releasing) 
2.6.x-based product and IMO the 2.6 kernel is still a moving target, 
with non-trivial behavior (and sometimes API) changes every couple of 
kernel versions.

	Jeff