Re: [PATCH net-next] net/packet: Add getsockopt support for PACKET_COPY_THRESH

[Jason Xing <kerneljasonxing@gmail.com>]

On Fri, Mar 8, 2024 at 7:43 PM Willem de Bruijn
<willemdebruijn.kernel@gmail.com> wrote:
>
> Juntong Deng wrote:
> > Currently getsockopt does not support PACKET_COPY_THRESH,
> > and we are unable to get the value of PACKET_COPY_THRESH
> > socket option through getsockopt.
> >
> > This patch adds getsockopt support for PACKET_COPY_THRESH.
> >
> > Signed-off-by: Juntong Deng <juntong.deng@outlook.com>
> > ---
> >  net/packet/af_packet.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
> > index 0db31ca4982d..65042edd1a97 100644
> > --- a/net/packet/af_packet.c
> > +++ b/net/packet/af_packet.c
> > @@ -4090,6 +4090,9 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
> >       case PACKET_VNET_HDR_SZ:
> >               val = READ_ONCE(po->vnet_hdr_sz);
> >               break;
> > +     case PACKET_COPY_THRESH:
> > +             val = pkt_sk(sk)->copy_thresh;
> > +             break;
>
> This is probably a good opportunity to use READ_ONCE/WRITE_ONCE for
> this variable that can be modified and read concurrently.
>
> Alternatively I can fix up all three locations in a follow-on patch.

Yes, variables which can be set through setsockopt related functions
have a chance to be modified when another process does read it.

The patch you mentioned goes like this, I think:

diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index 0db31ca4982d..b598dcafe441 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -2318,7 +2318,7 @@ static int tpacket_rcv(struct sk_buff *skb,
struct net_device *dev,
        }
        if (po->tp_version <= TPACKET_V2) {
                if (macoff + snaplen > po->rx_ring.frame_size) {
-                       if (po->copy_thresh &&
+                       if (READ_ONCE(po->copy_thresh) &&
                            atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
                                if (skb_shared(skb)) {
                                        copy_skb = skb_clone(skb, GFP_ATOMIC);
@@ -3836,7 +3836,7 @@ packet_setsockopt(struct socket *sock, int
level, int optname, sockptr_t optval,
                if (copy_from_sockptr(&val, optval, sizeof(val)))
                        return -EFAULT;

-               pkt_sk(sk)->copy_thresh = val;
+               WRITE_ONCE(pkt_sk(sk)->copy_thresh, val);
                return 0;
        }
        case PACKET_VERSION:
diff --git a/net/packet/diag.c b/net/packet/diag.c
index b3bd2f6c2bf7..47f69f3dbf73 100644
--- a/net/packet/diag.c
+++ b/net/packet/diag.c
@@ -17,7 +17,7 @@ static int pdiag_put_info(const struct packet_sock
*po, struct sk_buff *nlskb)
        pinfo.pdi_index = po->ifindex;
        pinfo.pdi_version = po->tp_version;
        pinfo.pdi_reserve = po->tp_reserve;
-       pinfo.pdi_copy_thresh = po->copy_thresh;
+       pinfo.pdi_copy_thresh = READ_ONCE(po->copy_thresh);
        pinfo.pdi_tstamp = READ_ONCE(po->tp_tstamp);

        pinfo.pdi_flags = 0;

Thanks,
Jason

>
> No concerns with adding the getsockopt itself.
>
> >       case PACKET_VERSION:
> >               val = po->tp_version;
> >               break;
> > --
> > 2.39.2
> >
>
>
>