[PATCH net-next v2 06/13] rxrpc: rxgk: Provide infrastructure and key derivation

David Howells dhowells at redhat.com
Wed Apr 9 23:54:09 PDT 2025


Jakub Kicinski <kuba at kernel.org> wrote:

> On Mon,  7 Apr 2025 17:11:19 +0100 David Howells wrote:
> > +	aead = crypto_krb5_prepare_encryption(krb5, &TK, RXGK_CLIENT_ENC_RESPONSE, gfp);
> > +	if (IS_ERR(aead))
> > +		goto aead_error;
> > +	gk->resp_enc = aead;
> > +
> > +	if (crypto_aead_blocksize(gk->resp_enc) != krb5->block_len ||
> > +	    crypto_aead_authsize(gk->resp_enc) != krb5->cksum_len) {
> > +		pr_notice("algo inconsistent with krb5 table %u!=%u or %u!=%u\n",
> > +			  crypto_aead_blocksize(gk->resp_enc), krb5->block_len,
> > +			  crypto_aead_authsize(gk->resp_enc), krb5->cksum_len);
> > +		return -EINVAL;
> 
> kfree_sensitive(buffer); missing?

Good catch, thanks.  That path should never trigger, but it should really do
"ret = -EINVAL; goto out;".

Do you want me to respin the patches or follow up with a fix patch?

David




More information about the linux-afs mailing list