Re: 2.6.18-rc1

[Nick Piggin <nickpiggin@yahoo.com.au>]

Con Kolivas wrote:
> I see the merge window closed and swap prefetch got bypassed again. I'd like 
> to believe it was an oversight but far more likely that Andrew remains 
> undecided about whether it should go in or not.
> 
> No bug reports have come from it in 6 months, the code has remained unchanged 
> for 3 months, it is as unobtrusive as a driver that is not compiled in 
> when !CONFIGed and there are numerous reports from satisfied users (even ones 
> that made it to the scary grounds of lkml). The only thing that happens is 
> Nick keeps threatening to review it over and over and over and....

I was going to review it again, but I noticed it has still has comments
(from Hugh and I, I believe) which still haven't been implemented. Like
duplicating most of read_swap_cache_async. I thought you might have some
improvements on the way, so I hadn't bothered yet.

But... excuse me? I *threaten* to review it? I volunteered to review it a
couple of times and found several problems. But OK if you take that as a
threat, then I won't review it.

And I haven't seen any numbers to show it even works in ideal conditions
after I told you how to fix the watermark code, let alone the real world
situations in which it is supposed to help (not that that seems to be a
showstopper to merging stuff like this, though)

I personally won't advocate it, but I wouldn't be upset if it goes
in... it isn't entirely unobtrusive: it is pretty close to the core mm,
and will have to be maintained as such. Mainly in Hugh's area, so he
would have a final veto there.

> 
> I'm not sure what else needs to happen?
> 

Probably if nothing happens, it sounds like Andrew will merge it
eventually.

-- 
SUSE Labs, Novell Inc.

Send instant messages to your online friends http://au.messenger.yahoo.com