Re: [PATCH] crypto: cavium - prevent integer overflow loading firmware

[Dan Carpenter <dan.carpenter@oracle.com>]

On Thu, Sep 08, 2022 at 05:42:54PM +0800, Herbert Xu wrote:
> On Thu, Sep 01, 2022 at 06:32:53PM +0300, Dan Carpenter wrote:
> >
> > @@ -263,7 +264,13 @@ static int cpt_ucode_load_fw(struct cpt_device *cpt, const u8 *fw, bool is_ae)
> >  	ucode = (struct ucode_header *)fw_entry->data;
> >  	mcode = &cpt->mcode[cpt->next_mc_idx];
> >  	memcpy(mcode->version, (u8 *)fw_entry->data, CPT_UCODE_VERSION_SZ);
> > -	mcode->code_size = ntohl(ucode->code_length) * 2;
> > +
> > +	code_length = ntohl(ucode->code_length);
> > +	if (code_length >= INT_MAX / 2) {
> > +		ret = -EINVAL;
> > +		goto fw_release;
> > +	}
> > +	mcode->code_size = code_length;
> 
> Where did the "* 2" go?

Crud.  :/  Sorry.

> 
> BTW, what is the threat model here? If the firmware metadata can't
> be trusted, shouldn't we be capping the firmware size at a level
> a lot lower than INT_MAX?

This is not firmware metadata, I'm fairly sure the fw_entry->data is raw
data from the file system.  Realistically, if you can't trust your
firmware then you are probably toasted but there is a move to trust as
little as possible.  Also Smatch marks data from the file system as
untrusted so it will generate static checker warnings.

regards,
dan carpenter