Re: [PATCH] lib/mpi: Fix buffer overrun when SG is too long

[Eric Biggers <ebiggers@kernel.org>]

On Tue, Dec 20, 2022 at 11:36:50AM +0100, Roberto Sassu wrote:
> On 12/20/2022 8:24 AM, Herbert Xu wrote:
> > On Mon, Dec 19, 2022 at 09:49:29AM +0100, Roberto Sassu wrote:
> > > 
> > > do you have any news on this bug?
> > 
> > Thanks for the reminder.  Could you please try this patch?
> 
> Tried, could not boot the UML kernel.
> 
> After looking, it seems we have to call sg_miter_stop(). Or alternatively,
> we could let sg_miter_next() be called but not writing anything inside the
> loop.
> 
> With either of those fixes, the tests pass (using one scatterlist).
> 
> Roberto
> 
> > ---8<---
> > The helper mpi_read_raw_from_sgl ignores the second parameter
> > nbytes when reading the SG list and may overrun its own buffer
> > because it only allocates enough memory according to nbytes.
> > 
> > Fixes: 2d4d1eea540b ("lib/mpi: Add mpi sgl helpers")
> > Reported-by: Roberto Sassu <roberto.sassu@huaweicloud.com>
> > Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
> > 
> > diff --git a/lib/mpi/mpicoder.c b/lib/mpi/mpicoder.c
> > index 39c4c6731094..6bffc68c1a5a 100644
> > --- a/lib/mpi/mpicoder.c
> > +++ b/lib/mpi/mpicoder.c
> > @@ -494,17 +494,15 @@ MPI mpi_read_raw_from_sgl(struct scatterlist *sgl, unsigned int nbytes)
> >   	val->sign = 0;
> >   	val->nlimbs = nlimbs;
> > -	if (nbytes == 0)
> > -		return val;
> > -
> >   	j = nlimbs - 1;
> >   	a = 0;
> >   	z = BYTES_PER_MPI_LIMB - nbytes % BYTES_PER_MPI_LIMB;
> >   	z %= BYTES_PER_MPI_LIMB;
> > -	while (sg_miter_next(&miter)) {
> > +	while (nbytes && sg_miter_next(&miter)) {
> >   		buff = miter.addr;
> > -		len = miter.length;
> > +		len = min_t(unsigned, miter.length, nbytes);
> > +		nbytes -= len;
> >   		for (x = 0; x < len; x++) {
> >   			a <<= 8;

I think it should look like:

	while (nbytes) {
		sg_miter_next(&miter);
		...
	}
	sg_miter_stop(&miter);

- Eric