From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jakub Kicinski <jakub.kicinski@netronome.com>
Subject: Re: [PATCH net-next v2 7/8] netdevsim: add SR-IOV functionality
Date: Fri, 1 Dec 2017 13:45:09 -0800
Message-ID: <20171201134509.2b1dc343@cakuba.netronome.com>
References: <20171201013540.18106-1-jakub.kicinski@netronome.com>
        <20171201013540.18106-8-jakub.kicinski@netronome.com>
        <20171201134306.GQ32305@orbyte.nwl.cc>
        <20171201121407.4f802f18@cakuba.netronome.com>
        <20171201213652.GT32305@orbyte.nwl.cc>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: netdev@vger.kernel.org, davem@davemloft.net, daniel@iogearbox.net,
        alexei.starovoitov@gmail.com, jiri@resnulli.us,
        oss-drivers@netronome.com, Sabrina Dubroca <sd@queasysnail.net>
To: Phil Sutter <phil@nwl.cc>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mail-qk0-f196.google.com ([209.85.220.196]:42522 "EHLO
        mail-qk0-f196.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1751031AbdLAVpN (ORCPT
        <rfc822;netdev@vger.kernel.org>); Fri, 1 Dec 2017 16:45:13 -0500
Received: by mail-qk0-f196.google.com with SMTP id a71so14941741qkc.9
        for <netdev@vger.kernel.org>; Fri, 01 Dec 2017 13:45:13 -0800 (PST)
In-Reply-To: <20171201213652.GT32305@orbyte.nwl.cc>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On Fri, 1 Dec 2017 22:36:52 +0100, Phil Sutter wrote:
> On Fri, Dec 01, 2017 at 12:14:07PM -0800, Jakub Kicinski wrote:
> > On Fri, 1 Dec 2017 14:43:06 +0100, Phil Sutter wrote:  
> > > On Thu, Nov 30, 2017 at 05:35:39PM -0800, Jakub Kicinski wrote:
> > > [...]  
> > > > +static int nsim_vfs_enable(struct netdevsim *ns, unsigned int num_vfs)
> > > > +{
> > > > +	ns->vfconfigs = kcalloc(num_vfs, sizeof(struct nsim_vf_config),
> > > > +				GFP_KERNEL);
> > > > +	if (!ns->vfconfigs)
> > > > +		return -ENOMEM;
> > > > +	ns->num_vfs = num_vfs;
> > > > +
> > > > +	return 0;
> > > > +}
> > > > +
> > > > +static void nsim_vfs_disable(struct netdevsim *ns)
> > > > +{
> > > > +	kfree(ns->vfconfigs);
> > > > +	ns->vfconfigs = NULL;
> > > > +	ns->num_vfs = 0;
> > > > +}    
> > > 
> > > Why not something like:
> > > 
> > > | static int nsim_vfs_set(struct netdevsim *ns, unsigned int num_vfs)
> > > | {
> > > | 	void *ptr = krealloc(ns->vfconfigs,
> > > | 			     num_vfs * sizeof(struct nsim_vf_config),
> > > | 			     GFP_KERNEL);
> > > | 
> > > | 	if (!ptr)
> > > | 		return -ENOMEM;
> > > | 
> > > | 	ns->vfconfigs = ptr;
> > > | 	ns->num_vfs = num_vfs;
> > > | 	return 0;
> > > | }  
> > 
> > Um.  It either frees or allocates, never reallocates so I felt realloc
> > is misleading.  ZERO_SIZE_PTR is less clearly a NULL than a NULL.  I
> > will have to specify __GFP_ZERO.  It's not a calloc so there could be
> > potentially some overflows?  
> 
> I don't understand: How can overflows happen if I use malloc() instead
> of calloc()?

The multiplication may overflow.  That's why we have kmalloc_array().
Note this explicit check in kmalloc_array() (which is also called by
kcalloc):

	if (size != 0 && n > SIZE_MAX / size)
		return NULL;

Where:

#define SIZE_MAX	(~(size_t)0)

> > > > +	ret = count;
> > > > +exit_unlock:
> > > > +	rtnl_unlock();
> > > > +
> > > > +	return ret;
> > > > +}    
> > > 
> > > [...]
> > >   
> > > > +static void nsim_free(struct net_device *dev)
> > > > +{
> > > > +	struct netdevsim *ns = netdev_priv(dev);
> > > > +
> > > > +	device_unregister(&ns->dev);
> > > >  }    
> > > 
> > > Shouldn't this also kfree(ns->vfconfigs)?  
> > 
> > It's in uninit, I will move it to release.  
> 
> Oh, I missed that. If you're certain this won't lead to memleaks, no
> objection from my side. :)

OK, I will respin v3 with the free moved :)