[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20180518205011.GA9756@avx2>
Date: Fri, 18 May 2018 23:50:11 +0300
From: Alexey Dobriyan <adobriyan@...il.com>
To: Al Viro <viro@...IV.linux.org.uk>
Cc: linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
hch@....de
Subject: Re: [PATCH -vfs] proc: don't maintain sizeof(struct proc_dir_entry)
On Fri, May 18, 2018 at 09:23:15PM +0100, Al Viro wrote:
> On Fri, May 18, 2018 at 11:02:13PM +0300, Alexey Dobriyan wrote:
> > Automatically cap sizeof(struct proc_dir_entry) at 192/128 bytes or
> > 256/192 bytes if spinlock debugging/lockdep is enabled.
>
> > --- a/fs/proc/internal.h
> > +++ b/fs/proc/internal.h
> > @@ -61,13 +61,15 @@ struct proc_dir_entry {
> > char *name;
> > umode_t mode;
> > u8 namelen;
> > + char inline_name[];
> > +} __randomize_layout;
> > +
>
> Hmm... proc_root initialization relies upon a gccism after that...
Yeah well.
> Speaking of which, why the hell do we even bother with proc_root.name?
printk("%s/%s", pde->parent->name, pde->name) is nice and doesn't cost
anything (OK, it may cost 5 bytes).
Powered by blists - more mailing lists