Re: [PATCH 08/11] NFS: Introduce lifecycle management for labelattribute.

[Dave Quigley]

On Thu, 2008-02-28 at 12:04 +1100, James Morris wrote:
> On Wed, 27 Feb 2008, David P. Quigley wrote:
> 
> > +#ifdef CONFIG_SECURITY
> > +static inline void nfs_fattr_alloc(struct nfs_fattr *fattr, gfp_t flags)
> > +{
> > +	fattr->label = kzalloc(NFS4_MAXLABELLEN, flags);
> > +	if (fattr->label == NULL)
> > +		panic("Can't allocate security label.");
> > +	fattr->label_len = NFS4_MAXLABELLEN;
> > +}
> 
> A panic here seems like overkill, and also possibly a DoS vector.  I 
> suggest having the calling code handle the allocation failure gracefully.
> 

Good point, I'll put this on the list. I think I remember you mentioned
something about this before but it must have slipped through the cracks.

> > +
> > +#define	nfs_fattr_fini(fattr)	_nfs_fattr_fini(fattr, __FILE__, __LINE__, __func__)
> > +static inline void _nfs_fattr_fini(struct nfs_fattr *fattr,
> > +		const char *file, int line, const char *func)
> > +{
> > +	if ((fattr)->label == NULL) {
> > +		if (fattr->label_len != 0) {
> > +			printk(KERN_WARNING
> > +					"%s:%d %s() nfs_fattr label available (%d)\n",
> > +					file, line, func,
> > +					fattr->label_len);
> > +		}
> > +	} else {
> > +		if (fattr->label_len == NFS4_MAXLABELLEN)
> > +			printk(KERN_WARNING
> > +					"%s:%d %s() nfs_fattr label unused\n",
> > +					file, line, func);
> > +		else if (fattr->label_len != (strlen(fattr->label) + 1))
> > +			printk(KERN_WARNING
> > +				"%s:%d %s() nfs_fattr label size mismatch (label_len %d, strlen %d)\n",
> > +				file, line, func,
> > +				fattr->label_len, strlen(fattr->label) + 1);
> > +
> > +		kfree(fattr->label);
> > +		fattr->label = NULL;
> > +		fattr->label_len = 0;
> > +	}
> > +}
> > +#else
> > +#define	nfs_fattr_alloc(fattr, flags)
> > +#define	nfs_fattr_fini(fattr)
> > +#endif
> 
> Perhaps introduce a debug configuration option for this code.

In what way? We can change them to respond to the NFS/NFSD debug
variables that are set through proc, or did you have something else in
mind?

> 
> 
> - James

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/