Re: [PATCH] net: clear heap allocations for privileged ethtool actions

[Ben Hutchings <bhutchings@solarflare.com>]

On Thu, 2010-10-07 at 23:31 +0200, Eric Dumazet wrote:
> Le jeudi 07 octobre 2010 à 14:10 -0700, Kees Cook a écrit :
> > Several other ethtool functions leave heap uncleared (potentially) by
> > drivers. Some interfaces appear safe (eeprom, etc), in that the sizes
> > are well controlled. In some situations (e.g. unchecked error conditions),
> > the heap will remain unchanged in areas before copying back to userspace.
> > Note that these are less of an issue since these all require CAP_NET_ADMIN.
> 
> > @@ -775,7 +775,7 @@ static int ethtool_get_regs(struct net_device *dev, char __user *useraddr)
> >  	if (regs.len > reglen)
> >  		regs.len = reglen;
> >  
> > -	regbuf = kmalloc(reglen, GFP_USER);
> > +	regbuf = kzalloc(reglen, GFP_USER);

Actually, I recently changed this to vmalloc() so your patch won't
apply.

> >  	if (!regbuf)
> >  		return -ENOMEM;
> >  
> > -- 
> > 1.7.1
> > 
> 
> Are you sure this is not hiding a more problematic problem ?
> 
> Code does :
> 
> reglen = ops->get_regs_len(dev);
> if (regs.len > reglen)
> 	regs.len = reglen;
> regbuf = kmalloc(reglen, GFP_USER);
> 
> So we can not copy back kernel memory.
> 
> However, what happens if user provides regs.len = 1 byte, and driver
> get_regs() doesnt properly checks regs.len and write past end of regbuf
> -> We probably write on other parts of kernel memory
[...]

Why should the driver's get_regs() check regs.len?  The buffer is
allocated based on reglen which is provided by the driver, not the user.

reglen (length of the kernel buffer) is not reduced; regs.len (length of
the user buffer) is.  That lets the user know how much of the user
buffer was actually used.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.