From mboxrd@z Thu Jan  1 00:00:00 1970
From: Alexey Dobriyan <adobriyan@sw.ru>
Subject: Re: [Devel] Re: [PATCH resend] proc: Fix proc_kill_inodes to kill dentries on all proc superblocks
Date: Thu, 1 Nov 2007 18:59:38 +0300
Message-ID: <20071101155938.GB16499@localhost.sw.ru>
References: <m1ir4mezzy.fsf@ebiederm.dsl.xmission.com> <4729F549.6090905@openvz.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Return-path: <linux-kernel-owner+glk-linux-kernel-3=40m.gmane.org-S1757577AbXKAQBN@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <4729F549.6090905@openvz.org>
Sender: linux-kernel-owner@vger.kernel.org
To: Pavel Emelyanov <xemul@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>, Andrew Morton <akpm@linux-foundation.org>, Linux Containers <containers@lists.osdl.org>, linux-kernel@vger.kernel.org
List-Id: containers.vger.kernel.org

On Thu, Nov 01, 2007 at 06:48:25PM +0300, Pavel Emelyanov wrote:
> Eric W. Biederman wrote:
> > It appears we overlooked support for removing generic proc files
> > when we added support for multiple proc super blocks.  Handle
> > that now.
> > 
> > Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
> 
> AFAIS this is just making the kill for all the super blocks
> we have.
> 
> Acked-by: Pavel Emelyanov <xemul@openvz.org>

bzzrt, this function is executed only for directories and symlinks,
obviously buggy (you can't flip ->f_op on live file) and is going away
after readdir/rmmod races for directories will be dealt the same way as
for regular file.

> > --- a/fs/proc/generic.c
> > +++ b/fs/proc/generic.c
> > @@ -561,28 +561,32 @@ static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp
> >  static void proc_kill_inodes(struct proc_dir_entry *de)
> >  {
> >  	struct list_head *p;
> > -	struct super_block *sb = proc_mnt->mnt_sb;
> > +	struct super_block *sb;
> >  
> >  	/*
> >  	 * Actually it's a partial revoke().
> >  	 */
> > -	file_list_lock();
> > -	list_for_each(p, &sb->s_files) {
> > -		struct file * filp = list_entry(p, struct file, f_u.fu_list);
> > -		struct dentry * dentry = filp->f_path.dentry;
> > -		struct inode * inode;
> > -		const struct file_operations *fops;
> > -
> > -		if (dentry->d_op != &proc_dentry_operations)
> > -			continue;
> > -		inode = dentry->d_inode;
> > -		if (PDE(inode) != de)
> > -			continue;
> > -		fops = filp->f_op;
> > -		filp->f_op = NULL;
> > -		fops_put(fops);
> > +	spin_lock(&sb_lock);
> > +	list_for_each_entry(sb, &proc_fs_type.fs_supers, s_instances) {
> > +		file_list_lock();
> > +		list_for_each(p, &sb->s_files) {
> > +			struct file * filp = list_entry(p, struct file, f_u.fu_list);
> > +			struct dentry * dentry = filp->f_path.dentry;
> > +			struct inode * inode;
> > +			const struct file_operations *fops;
> > +
> > +			if (dentry->d_op != &proc_dentry_operations)
> > +				continue;
> > +			inode = dentry->d_inode;
> > +			if (PDE(inode) != de)
> > +				continue;
> > +			fops = filp->f_op;
> > +			filp->f_op = NULL;
> > +			fops_put(fops);
> > +		}
> > +		file_list_unlock();
> >  	}
> > -	file_list_unlock();
> > +	spin_unlock(&sb_lock);
> >  }

Yup, all this bogo-revoke is going away RSN. you just kicked my lazy ass.