Re: Major regression on hackbench with SLUB (more numbers)

[Al Viro <viro@ftp.linux.org.uk>]

On Sat, Dec 22, 2007 at 09:50:09PM +0000, Al Viro wrote:
> On Sat, Dec 22, 2007 at 01:00:09PM -0800, Linus Torvalds wrote:
> 
> > > Another problem with using /sys/slab is that it is downright *ugly*.
> > > Why is it for example, that /sys/slab/dentry is a symlink to
> > > ../slab/:a-0000160?
> > 
> > That's the only really ugly thing there. Otherwise, it's pretty nice, but 
> > having a million files makes for problems when trying to send somebody 
> > else the full info. 
> 
> *raised brows*
> 
> I would say that there's that really ugly thing with embedding kobject
> into a struct with the lifetime rules of its own and then having that
> struct kfree()d while kobject might still have references, but hey,
> maybe it's just me and my lack of appreciation of the glory that is
> sysfs.
> 
> 	Al, too tired of ranting about sysfs being a major architecture
> mistake and a recurring source of turds like that one...

BTW, the fact that presence of that kobject is conditional makes life even
uglier - we have to either kfree() or drop a reference to kobject leaving
actual kfree() to its ->release().   While we are at it, when do we remove
the symlinks?  That got to add even more fun for the lifetime rules...

Sigh...  How does one set a script that would mail a warning upon git pull
that introduces any instances of keyword from given set?  I hadn't noticed
that slub was using sysfs when it went into the tree back in May ;-/