Re: [PATCH net-next v2 7/8] netdevsim: add SR-IOV functionality

[Phil Sutter <phil@nwl.cc>]

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()?

> > > +static ssize_t
> > > +nsim_numvfs_store(struct device *dev, struct device_attribute *attr,
> > > +		  const char *buf, size_t count)
> > > +{
> > > +	struct netdevsim *ns = to_nsim(dev);
> > > +	unsigned int num_vfs;
> > > +	int ret;
> > > +
> > > +	ret = kstrtouint(buf, 0, &num_vfs);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	rtnl_lock();
> > > +	if (ns->num_vfs == num_vfs)
> > > +		goto exit_good;  
> > 
> > Then replace this:
> > 
> > > +	if (ns->num_vfs && num_vfs) {
> > > +		ret = -EBUSY;
> > > +		goto exit_unlock;
> > > +	}
> > > +
> > > +	if (num_vfs) {
> > > +		ret = nsim_vfs_enable(ns, num_vfs);
> > > +		if (ret)
> > > +			goto exit_unlock;
> > > +	} else {
> > > +		nsim_vfs_disable(ns);
> > > +	}  
> > 
> > with just:
> > 
> > |	nsim_vfs_set(ns, num_vfs);
> 
> I'm trying to mirror the PCI subsystem behaviour here, which only
> allows enable or disable, not increase.  I felt we should follow how
> real devices behave:
> 
> 	/* enable VFs */
> 	if (pdev->sriov->num_VFs) {
> 		dev_warn(&pdev->dev, "%d VFs already enabled. Disable before enabling %d VFs\n",
> 			 pdev->sriov->num_VFs, num_vfs);
> 		return -EBUSY;
> 	}
> 
> So IOW this is intentional.

Ah, I see. Yes, then it makes sense! Keeping this virtual VF
functionality as close to real ones as possible is certainly feasible.

> > > +	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. :)

Cheers, Phil