On Fri, Jan 17, 2020 at 03:47:20PM -0600, Shiraz Saleem wrote:
> From: "Shiraz Saleem" <shiraz.saleem@xxxxxxxxx>
>
> The in_dev_for_each_ifa_rtnl iterator in i40iw_add_ipv4_addr
> requires that the rtnl lock be held. But the rtnl_trylock/unlock
> scheme in this function does not guarantee it.
>
> Replace the rtnl locking with an RCU lookup since there are
> no netdev object updates in this function.
>
> Fixes: 8e06af711bf2 ("i40iw: add main, hdr, status")
> Signed-off-by: Shiraz Saleem <shiraz.saleem@xxxxxxxxx>
> ---
> drivers/infiniband/hw/i40iw/i40iw_main.c | 16 +++++-----------
> 1 file changed, 5 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/infiniband/hw/i40iw/i40iw_main.c b/drivers/infiniband/hw/i40iw/i40iw_main.c
> index 2386143..d7a1219 100644
> --- a/drivers/infiniband/hw/i40iw/i40iw_main.c
> +++ b/drivers/infiniband/hw/i40iw/i40iw_main.c
> @@ -1212,22 +1212,19 @@ static void i40iw_add_ipv4_addr(struct i40iw_device *iwdev)
> {
> struct net_device *dev;
> struct in_device *idev;
> - bool got_lock = true;
> u32 ip_addr;
>
> - if (!rtnl_trylock())
> - got_lock = false;
> -
> - for_each_netdev(&init_net, dev) {
> + rcu_read_lock();
> + for_each_netdev_rcu(&init_net, dev) {
> if ((((rdma_vlan_dev_vlan_id(dev) < 0xFFFF) &&
> (rdma_vlan_dev_real_dev(dev) == iwdev->netdev)) ||
> (dev == iwdev->netdev)) && (dev->flags & IFF_UP)) {
> const struct in_ifaddr *ifa;
The rtnl does not just protect the pointers, but also values like
flags from changing. When using RCU any varient values of the dev
should be read with READ_ONCE, such as flags.
I'm not completely sure the vlan tests are OK under RCU either, but
netdev has historically been loose on its READ_ONCE/WRITE_ONCE
annotations..
Jason
