Em Wed, Feb 27, 2008 at 08:36:50AM +0100, Ingo Molnar escreveu:

* David Miller [off-list ref] wrote:

quoted

quoted

but the main and fundamental question still remains unanswered (more 
than 3 weeks after Andrew asked that question): why was this piece 
of general infrastructure merged via net.git and not submitted to 
lkml ever? The code touching -mm does _not_ count as "review".

I already stated this was a mistake and it won't happen again in the 
future.

sorry - that bit of the thread didnt seem to make it to lkml. I just saw 
this incomplete discussion with a denial and with no resolution.

And you did the right thing anyway by thinking in terms of a generic 
piece of infrastructure instead of hiding it away into say 
include/net/pcounter.h (which nobody could have objected against).

I sometimes think that the forced isolation of subsystems (rather 
strongly enforced both by -mm and by linux-next) and their hiding away 
on non-lkml lists will eventually hurt the core kernel because less and 
less people will be willing to go the trouble of doing proper 
cross-subsystem development. That results in duplicated or specialistic 
infrastructure, increased code size and longer term, ultimately less 
performance. (by the time we notice _that_ it will probably be too late 
to do anything about it)

This more friendly wording makes me feel actually happy to get from my
hiding place and tell that I actually saw the percpu counters code, just
after I made Eric's code generic as I thought it should have been from
the start, I found out about what was already in lib/.

I just got lazy to do what at the time looked the right thing
to do: to read thru the existing lib/ code and use it where pcounter was
being used. But at least it got exposed 8-)

Anyway, progress was made, I do not feel too bad about it even now.

- Arnaldo