Re: [Patch] Convert max_user_watches to long.

[Davide Libenzi]

On Wed, 27 Oct 2010, Andrew Morton wrote:

> On Wed, 27 Oct 2010 14:09:15 -0500
> Robin Holt <holt@xxxxxxx> wrote:
> 
> > On a 16TB machine, max_user_watches has an integer overflow.  Convert it
> > to use a long and handle the associated fallout.
> > 
> 
> hand-wavy reality check:
> 
> Are the existing defaults sane?  How well does the code perform with a
> few billion watches?

Algorithmically, wake ups events are O(1) and add/mod/del are log2(N).
Tell you what though, if you give me a 16TB RAM box, I will tell you how 
it performs in reality ;)



> Is the expected use case one-watch-per-user-per-fd?  If so, then
> perhaps the max number of user_watches should have some realtionship
> with the max number of fds?

Yes, the expected case is one watch per fd, but the reason the watch limit 
went in in the first place, was because of DoS potential of someone not 
playing nicely.



- Davide


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/