Re: [Lse-tech] Re: RFC: patch to allow lock-free traversal of lists with insertion

[Alexander V. <vi...@ma...>]

On 13 Oct 2001, Andi Kleen wrote:

> In article <Pin...@pe...>,
> Linus Torvalds <tor...@tr...> writes:
> 
> >  - nobody has shown a case where existing normal locking ends up being
> >    really a huge problem, and where RCU clearly helps.
> 
> The poster child of such a case is module unloading. Keeping reference
> counts for every even non sleeping use of a module is very painful. 
> The current "fix" -- putting module count increases in all possible module 
> callers to fix the unload races is slow and ugly and far too subtle to 
> get everything right. Waiting quiescent periods before unloading is a nice 
> alternative.

... while quiescent stuff is _not_ subtle and not prone to breakage.  Right.
In the same world where Vomit-Making System is elegant, SGI "designs" are
and NT is The Wave Of Future(tm).  Pardon me, but I'll stay in our universe
and away from the drugs of such power.