Re: [RFC PATCH v4] akcipher: new verify API for public key algorithms

[Vitaly Chikunov <vt@altlinux.org>]

Herbert,

On Fri, Feb 01, 2019 at 02:26:55PM +0800, Herbert Xu wrote:
> On Fri, Jan 25, 2019 at 09:01:16PM +0300, Vitaly Chikunov wrote:
> >
> > @@ -781,36 +780,17 @@ static int tpm_key_verify_signature(const struct key *key,
> >  	if (!req)
> >  		goto error_free_tfm;
> >  
> > -	ret = -ENOMEM;
> > -	outlen = crypto_akcipher_maxsize(tfm);
> > -	output = kmalloc(outlen, GFP_KERNEL);
> > -	if (!output)
> > -		goto error_free_req;
> > -
> > -	sg_init_one(&sig_sg, sig->s, sig->s_size);
> > -	sg_init_one(&digest_sg, output, outlen);
> > -	akcipher_request_set_crypt(req, &sig_sg, &digest_sg, sig->s_size,
> > -				   outlen);
> > +	sg_init_table(&src_sg, 2);
> > +	sg_set_buf(&src_sg[0], sig->s, sig->s_size);
> > +	sg_set_buf(&src_sg[1], sig->digest, sig->digest_size);
> > +	akcipher_request_set_crypt(req, &src_sg, NULL, sig->s_size,
> > +				   sig->digest_size);
> 
> It's not clear that sig->digest is guaranteed to be kmalloc memory.
> In any case, it's best not to mix unrelated changes in a single
> patch.  So please keep the kmalloc on output and then copy
> sig->digest into it and put output into the SG list.

It is not guaranteed that sig->s will be kmalloc memory either. (Except
we know it for sure like we know the same about sig->digest).

You can see in public_key_signature_free() that both fields are kfree'd
together.

So, I don't understand why we should treat sig->digest differently than
sig->s.

I was just removing kmalloc'ed output as crypto_akcipher_verify() does
not need any output anymore. So, it's not some sort of mixing unrelated
changes, from my point of view.

Thanks,

> 
> Thanks,
> -- 
> Email: Herbert Xu <herbert@gondor.apana.org.au>
> Home Page: http://gondor.apana.org.au/~herbert/
> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt