On Tue, Mar 13, 2018 at 07:03:54PM +0200, Yuval Shaia wrote:
> On Tue, Mar 13, 2018 at 04:06:14PM +0200, Leon Romanovsky wrote:
> > From: Parav Pandit <parav@xxxxxxxxxxxx>
> >
> > Currently initialized ah_attr contains the port number to which cm_id is
> > bound. However, while searching for GID table for matching GID entry,
> > port number is ignored.
> > It is better to honor the port number of the cm_id while searching GID
> > table for IB link layer as well.
> >
> > Reviewed-by: Daniel Jurgens <danielj@xxxxxxxxxxxx>
> > Signed-off-by: Parav Pandit <parav@xxxxxxxxxxxx>
> > Signed-off-by: Leon Romanovsky <leon@xxxxxxxxxx>
> > drivers/infiniband/core/multicast.c | 26 ++++++++++++--------------
> > 1 file changed, 12 insertions(+), 14 deletions(-)
> >
> > diff --git a/drivers/infiniband/core/multicast.c b/drivers/infiniband/core/multicast.c
> > index 45f2f095f793..4eb72ff539fc 100644
> > +++ b/drivers/infiniband/core/multicast.c
> > @@ -724,21 +724,19 @@ int ib_init_ah_from_mcmember(struct ib_device *device, u8 port_num,
> > {
> > int ret;
> > u16 gid_index;
> > - u8 p;
> > -
> > - if (rdma_protocol_roce(device, port_num)) {
> > - ret = ib_find_cached_gid_by_port(device, &rec->port_gid,
> > - gid_type, port_num,
> > - ndev,
> > - &gid_index);
> > - } else if (rdma_protocol_ib(device, port_num)) {
> > - ret = ib_find_cached_gid(device, &rec->port_gid,
> > - IB_GID_TYPE_IB, NULL, &p,
> > - &gid_index);
> > - } else {
> > - ret = -EINVAL;
> > - }
> >
> > + /* GID table is not based on the netdevice for IB link layer,
> > + * so ignore ndev during search.
> > + */
> > + if (rdma_protocol_ib(device, port_num))
> > + ndev = NULL;
> > + else if (!rdma_protocol_roce(device, port_num))
> > + return -EINVAL;
> > +
> > + ret = ib_find_cached_gid_by_port(device, &rec->port_gid,
> > + gid_type, port_num,
> > + ndev,
> > + &gid_index);
>
> It is unclear from commit message why for IB we are no longer using
> ib_find_cached_gid().
You mean vs ib_find_cached_gid_by_port()? Isn't that explained by
this:
It is better to honor the port number of the cm_id while searching GID
table for IB link layer as well.
'better' is a bit of an understatement..
If the the AH is supposed to be used on a specific port then it must
be used with the gid_index for the port and not accidently get
switched over to the gid_index for some other port.
So the gid must be present on the provided port or
ib_init_ah_from_mcmember must fail.. It was a bug to not check it.
Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
