RDMA/addr: Mark addr_resolve as might_sleep()

Under one path through ib_nl_fetch_ha() this calls nlmsg_new(GFP_KERNEL)
which is a sleeping call. This is a very rare path, so mark fetch_ha() and
the module external entry point that conditionally calls through to
fetch_ha() as might_sleep().

Link: https://lore.kernel.org/r/20200506074701.9775-2-leon@kernel.orgSigned-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

drivers/infiniband/core/addr.c

@@ -371,6 +371,8 @@ static int fetch_ha(const struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
 		(const void *)&dst_in6->sin6_addr;
 	sa_family_t family = dst_in->sa_family;
 
+	might_sleep();
+
 	/* If we have a gateway in IB mode then it must be an IB network */
 	if (has_gateway(dst, family) && dev_addr->network == RDMA_NETWORK_IB)
 		return ib_nl_fetch_ha(dev_addr, daddr, seq, family);
@@ -727,6 +729,8 @@ int roce_resolve_route_from_path(struct sa_path_rec *rec,
 	struct rdma_dev_addr dev_addr = {};
 	int ret;
 
+	might_sleep();
+
 	if (rec->roce.route_resolved)
 		return 0;