On Tue, 17 Mar 2009, Nick Piggin wrote:

> > Yes, my patch isn't realy solusion.

> > Andrea already pointed out that it's not O_DIRECT issue, it's gup vs fork

> > issue. *and* my patch is crazy slow :)

> 

> Well, it's an interesting question. I'd say it probably is more than

> just O_DIRECT. vmsplice too, for example (which I think is much harder

> to fix this way because the pages are retired by the other end of

> the pipe, so I don't think you can hold a lock across it).



Well, only the "fork()" has the race problem.



So having a fork-specific lock (but not naming it by directio) actually 

does make sense. The fork is much less performance-critical than most 

random mmap_sem users - and doesn't have the same scalability issues 

either (ie people probably _do_ want to do mmap/munmap/brk concurrently 

with gup lookup, but there's much less worry about concurrent fork() 

performance).



It doesn't necessarily make the general problem go away, but it makes the 

_particular_ race between get_user_pages() and fork() go away. Then you 

can do per-page flags or whatever and not have to worry about concurrent 

lookups.



			Linus

> splice() already has a callback for releasing the pages, so it's doable.

doable, maybe.

・ last resort
最後｛さいご｝の手段｛しゅだん｝、苦肉｛くにく｝の策、伝家の宝刀、頼みの綱、最終審｛さいしゅうしん｝
・I've got to go with the last resort. : こうなったら最後の手段しかない

On Wed, 11 Mar 2009 01:55:42 +0900 (JST)
Ryusuke Konishi wrote:

> I've been working for the past serveral months to take review comments
> and to continually solve users' problems come up in mainling list

Yes, the maintenance has been impressive.

> (thanks for all giving comments and feedbacks!). Also, I've tried to
> stabilize API and disk format to restrict additional changes and
> ensure backward compatibility.

Well. From the point of view of mainline linux, there is no
back-compatibility issue, because the fs hasn't been merged yet.

You perhaps have back-compatibility concerns for existing users of the
out-of-tree patch, but I'd encourage you to not worry about that too
much - there will be fairly few users and they are probably pretty
technical and will be able to cope with a migration. It's a _bit_ hard
on them but on the other hand, omitting back-compatibility code leads
to a better implementation for the long term.

What you should be more concerned about is forward-compatibility. What
arrangements do you presently have in place to be able to later alter the
on-disk format without causing too much disruption? Having a strong
design here will make changes easier to do and will lead to a better
filesystem.

Also.. Don't get _too_ concerned about freezing the on-disk format at
this time. You could put in a mount-time printk("the nilfs on-disk
format may change at any time - do not place critical data on a nilfs
filesystem") and we leave that in place for a few months while things
stabilise.

And yes, I was planning on sending nilfs in to Linus for 2.6.30 unless
someone has decent-sounding reasons to hold it back.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/