Re: [patch 5/6] vfs: optimization to /proc/<pid>/mountinfo patch

[Miklos Szeredi <miklos@szeredi.hu>]

> On Thu, Mar 13, 2008 at 10:26:46PM +0100, Miklos Szeredi wrote:
> > +static int prepend(char **buffer, int *buflen, const char *str,
> > +			  int namelen)
> 
> inline, please.

Akpm wrote earlier about this function:
| Both the newly-added inlines in this patch are wrong.  They will result in
| a larger and slower kernel.  This should be very well known by now.

Please fight that out with him :)

> > @@ -1911,10 +1902,9 @@ char *dentry_path(struct dentry *dentry,
> >  
> >  	spin_lock(&dcache_lock);
> >  	prepend(&end, &buflen, "\0", 1);
> > -	if (!IS_ROOT(dentry) && d_unhashed(dentry)) {
> > -		if (prepend(&end, &buflen, "//deleted", 9))
> > +	if (!IS_ROOT(dentry) && d_unhashed(dentry) &&
> > +		(prepend(&end, &buflen, " (deleted)", 10) != 0))
> >  			goto Elong;
> 
> That's a bad idea:
> 	* we bloody well might want to use it outside of procfs
> 	* //deleted is _better_; you can't have an empty path component,
> but you can have a pathname ending on " (deleted)".

This was again Andrew's comment (making it look the same as the links
in /proc/PID/fd), but here I have to agree that //deleted is probably
better in this case.

Miklos