> Various signedness issues and integer overflows have been fixed within
> kNFSd and the XDR decode functions of kernel 2.6.
> These bugs can be triggered remotely by sending a package with a trusted
> source IP address and a write request with a size greater then 2^31.
> The result will be a kernel Oops, it is unknown if this bug is otherwise
> exploitable yet.
> Kernel 2.4 nfsd code is different but may suffer from the same
> vulnerability.
>
>
The iSEC people have read the nfsd code from 2.4 and it seems to be
vulnerable too, however only authenticated clients could reach the
problematic places at all. Having a writeable NFS share is probably a
bad idea anyway...
> Various signedness issues and integer overflows have been fixed within
> kNFSd and the XDR decode functions of kernel 2.6.
> These bugs can be triggered remotely by sending a package with a trusted
> source IP address and a write request with a size greater then 2^31.
> The result will be a kernel Oops, it is unknown if this bug is otherwise
> exploitable yet.
> Kernel 2.4 nfsd code is different but may suffer from the same
> vulnerability.
>
>
The iSEC people have read the nfsd code from 2.4 and it seems to be
vulnerable too, however only authenticated clients could reach the
problematic places at all. Having a writeable NFS share is probably a
bad idea anyway...
[ reply ]