[lkml]   [2015]   [Feb]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [GIT PULL] please pull file-locking related changes for v3.20
On Mon, 16 Feb 2015 16:21:30 -0800
Linus Torvalds <> wrote:

> On Mon, Feb 16, 2015 at 4:02 PM, Jeff Layton <> wrote:
> >
> > Now that I look, it may be best to just revert this whole set for now.
> > Linus, are you amenable to doing that?
> Sure. But I'd prefer seeing how hard it would be to fix things first.
> If this was at the end of the release cycle, I'd revert it
> immediately. As it is, try to see how had it is.

Fair enough. I just didn't want to hold up -rc1. I should be able to
fix up the bugs within the next day or so.

I've got a small stack of fixes that I'll send along soon.

> The bugs I found might be as easy as just the attached (TOTALLY
> UNTESTED) patch. The comment about a higher-priority process and
> sched_resced() is just complete and utter crap. If somebody holds a
> read lock and upgrades it to a write lock, there is absolutely *zero*
> reason to let some higher-priority process get the write-lock first -
> that's just simply semantically wrong bullshit. "Higher priority" does
> not mean "can punch through locks".

The patch is pretty close to one that I have, so yes I think that will fix
it. There is one bug in the first loop though -- "old_fl" should be set
to "fl" there.

I'm also happy to remove the "drop the spinlock" thing. It's bothered
me for a while...

> Removing the silly incorrect counts should be trivial too. There
> really are not very many users, and they can just walk the list
> instead.

Yes, that's a straightforward revert.

> Now, if you've found other more fundamental bugs that look unfixable,
> then that might mean that reverting it all is unavoidable, but..
> Linus

No, I don't think there's anything unfixable there. I did find another
bug, but it's simple to fix.

Jeff Layton <>

 \ /
  Last update: 2015-02-17 02:01    [W:0.040 / U:3.060 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site