On Wed, Dec 18, 2013 at 02:40:34PM -0500, Bruce Fields wrote:
> On Mon, Nov 18, 2013 at 08:32:43AM -0800, Greg KH wrote:
> > On Wed, Nov 13, 2013 at 10:16:55AM -0500, Bruce Fields wrote:
> > > (Argh, sorry, with the right stable address cc'd this time I hope.)
> > >
> > > On Wed, Nov 06, 2013 at 0
On Mon, Nov 18, 2013 at 08:32:43AM -0800, Greg KH wrote:
> On Wed, Nov 13, 2013 at 10:16:55AM -0500, Bruce Fields wrote:
> > (Argh, sorry, with the right stable address cc'd this time I hope.)
> >
> > On Wed, Nov 06, 2013 at 03:10:04PM +, Al Viro wrote:
> > > FWIW, not taking ->i_lock there de
(Argh, sorry, with the right stable address cc'd this time I hope.)
On Wed, Nov 06, 2013 at 03:10:04PM +, Al Viro wrote:
> FWIW, not taking ->i_lock there definitely looks like a good thing. As for
> 64bit ->i_ino itself... Looks like the main problem is the shitload of
> printks - the actua
On Wed, Nov 06, 2013 at 03:10:04PM +, Al Viro wrote:
> FWIW, not taking ->i_lock there definitely looks like a good thing. As for
> 64bit ->i_ino itself... Looks like the main problem is the shitload of
> printks - the actual uses of ->i_ino are fine, but these suckers create
> a lot of noise
On Mon, Nov 04, 2013 at 12:53:00AM +, Al Viro wrote:
> Maybe... OTOH, that crap really needs doing something only with nfsd on
> filesystems with 64bit inode numbers living on 32bit hosts (i_ino is
> unsigned long, not u32 right now). Hell knows; I'm somewhat concerned about
> setups like e.
On Sun, Nov 03, 2013 at 03:39:14PM -0800, Linus Torvalds wrote:
> On Sun, Nov 3, 2013 at 11:54 AM, Al Viro wrote:
> >
> > IIRC, at some point such an attempt has seriously hurt iget() on 32bit
> > boxen, so we ended up deciding not to go there. Had been years ago,
> > though...
>
> Yeah, I think
On Sun, Nov 03, 2013 at 03:39:14PM -0800, Linus Torvalds wrote:
> On Sun, Nov 3, 2013 at 11:54 AM, Al Viro wrote:
> >
> > IIRC, at some point such an attempt has seriously hurt iget() on 32bit
> > boxen, so we ended up deciding not to go there. Had been years ago,
> > though...
>
> Yeah, I think
On Sun, Nov 3, 2013 at 11:54 AM, Al Viro wrote:
>
> IIRC, at some point such an attempt has seriously hurt iget() on 32bit
> boxen, so we ended up deciding not to go there. Had been years ago,
> though...
Yeah, I think the circumstances have changed. 32-bit is less
important, and iget() is much
On Sun, Nov 03, 2013 at 10:25:32AM -0800, Linus Torvalds wrote:
> Ugh.
>
> This is too late since I want to do 3.12 today, and it seems to not be
> a regression - from what I can tell, the problem has always existed,
> no?
>
> So I'll consider this post-3.12. However, it also strikes me that we
>
Ugh.
This is too late since I want to do 3.12 today, and it seems to not be
a regression - from what I can tell, the problem has always existed,
no?
So I'll consider this post-3.12. However, it also strikes me that we
should just clean things up, and make "i_ino" be an u64, and be able
to do this
10 matches
Mail list logo
