linux-hardening - Re: [PATCH] IB/mlx4: Avoid field-overflowing memcpy()

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <202106171239.C425161E8@keescook>
Date:   Thu, 17 Jun 2021 12:46:43 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Leon Romanovsky <leon@...nel.org>
Cc:     Jack Morgenstein <jackm@...dia.com>,
        Jason Gunthorpe <jgg@...pe.ca>,
        Yishai Hadas <yishaih@...dia.com>,
        Doug Ledford <dledford@...hat.com>,
        linux-kernel@...r.kernel.org, linux-rdma@...r.kernel.org,
        linux-hardening@...r.kernel.org
Subject: Re: [PATCH] IB/mlx4: Avoid field-overflowing memcpy()

On Thu, Jun 17, 2021 at 10:24:58AM +0300, Leon Romanovsky wrote:
> On Wed, Jun 16, 2021 at 01:37:44PM -0700, Kees Cook wrote:
> > In preparation for FORTIFY_SOURCE performing compile-time and run-time
> > field bounds checking for memcpy(), memmove(), and memset(), avoid
> > intentionally writing across neighboring array fields.
> > 
> > Use the ether_addr_copy() helper instead, as already done for smac.
> > 
> > Signed-off-by: Kees Cook <keescook@...omium.org>
> > ---
> >  drivers/infiniband/hw/mlx4/qp.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
> > index 2ae22bf50016..4a2ef7daaded 100644
> > --- a/drivers/infiniband/hw/mlx4/qp.c
> > +++ b/drivers/infiniband/hw/mlx4/qp.c
> > @@ -3144,7 +3144,7 @@ static int build_mlx_header(struct mlx4_ib_qp *qp, const struct ib_ud_wr *wr,
> >  		mlx->sched_prio = cpu_to_be16(pcp);
> >  
> >  		ether_addr_copy(sqp->ud_header.eth.smac_h, ah->av.eth.s_mac);
> > -		memcpy(sqp->ud_header.eth.dmac_h, ah->av.eth.mac, 6);
> > +		ether_addr_copy(sqp->ud_header.eth.dmac_h, ah->av.eth.mac);
> >  		memcpy(&ctrl->srcrb_flags16[0], ah->av.eth.mac, 2);
> >  		memcpy(&ctrl->imm, ah->av.eth.mac + 2, 4);
> 
> I don't understand the last three lines. We are copying 6 bytes to
> ah->av.eth.mac and immediately after that overwriting them.

I'm not following (the memcpy() is replaced by ether_addr_copy()). I only
see ah->av.eth.mac being read from (not written to). And the destinations
are {s,d}mac_h:

ah->av.eth.s_mac   -> sqp->ud_header.eth.smac_h (s_mac to smac_h: 6 bytes)
ah->av.eth.mac     -> sqp->ud_header.eth.dmac_h (mac to dmac_h: 6 bytes)

after that I see:

ah->av.eth.mac     -> &ctrl->srcrb_flags16[0] (2 bytes)
ah->av.eth.mac + 2 -> ctrl->imm (4 bytes)

The last two copies mac again in pieces, but I don't know what any of
this is actually used for, which I what I assume you're asking about. :)

> Jack, 
> 
> Do you remember what you wanted to achieve in commit
> 6ee51a4e866b ("mlx4: Adjust QP1 multiplexing for RoCE/SRIOV")
> 
> Thanks

-Kees

-- 
Kees Cook