netdev - Re: [1/2] tun: Only free a netdev when all tun descriptors are closed

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20090418190937.GA7799@dhcp-1-124.tlv.redhat.com>
Date:	Sat, 18 Apr 2009 22:09:39 +0300
From:	"Michael S. Tsirkin" <m.s.tsirkin@...il.com>
To:	Herbert Xu <herbert@...dor.apana.org.au>
Cc:	"Eric W. Biederman" <ebiederm@...ssion.com>,
	Patrick McHardy <kaber@...sh.net>,
	Matias Zabaljauregui <zabaljauregui@...il.com>, odie@...aau.dk,
	Rusty Russell <rusty@...tcorp.com.au>, lguest@...abs.org,
	virtualization@...ts.osdl.org,
	"David S. Miller" <davem@...emloft.net>, netdev@...r.kernel.org,
	Christian Borntraeger <borntraeger@...ibm.com>
Subject: Re: [1/2] tun: Only free a netdev when all tun descriptors are
	closed

On Fri, Apr 17, 2009 at 08:09:23AM +0800, Herbert Xu wrote:
> On Thu, Apr 16, 2009 at 10:57:45PM +0300, Michael S. Tsirkin wrote:
> >
> > This last bit seems to make a simple test using a non-persistent tap device
> > deadlock for me: we don't drop a reference acquired with __tun_get sock
> > unregister_netdevice blocks printing unregister_netdevice: waiting for tap0 to
> > become free. Usage count = 1.
> 
> Ah yes, I'd overlooked the fact that the original code didn't
> require the tfile refcount to hit zero.  Now we do.  Here's an
> updated version of the first patch.  The second patch should still
> work as is.
> 
> tun: Only free a netdev when all tun descriptors are closed

...

> @@ -1275,20 +1278,17 @@ static int tun_chr_close(struct inode *inode, struct file *file)
>  	struct tun_file *tfile = file->private_data;
>  	struct tun_struct *tun = __tun_get(tfile);
>  
> -
>  	if (tun) {
> -		DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
> -
> -		rtnl_lock();
> -		__tun_detach(tun);
> -
>  		/* If desireable, unregister the netdevice. */
> -		if (!(tun->flags & TUN_PERSIST)) {
> -			sock_put(tun->sk);
> -			unregister_netdevice(tun->dev);
> -		}
> +		if (!(tun->flags & TUN_PERSIST))
> +			unregister_netdev(tun->dev);
> +		tun_put(tun);
> +	} else
> +		tun = tfile->tun;
>  
> -		rtnl_unlock();
> +	if (tun) {
> +		DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
> +		sock_put(tun->sk);
>  	}
>  
>  	put_net(tfile->net);

Does this work with TUN_PERSIST off?
I haven't tested this, but won't unregister_netdev block forever
waiting for device reference to become 0? Maybe you want

+		tun_put(tun);
+		if (!(tun->flags & TUN_PERSIST))
+			unregister_netdev(tun->dev);

or is there a race here?

-- 
MST
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html