在 2021/6/2 21:10, Kamal Heib 写道:
On Tue, Jun 01, 2021 at 02:01:32PM -0300, Jason Gunthorpe wrote:
On Tue, Jun 01, 2021 at 11:59:44PM +0800, Zhu Yanjun wrote:
On Tue, Jun 1, 2021 at 5:09 PM Kamal Heib <kamalheib1@xxxxxxxxx> wrote:
On Tue, Jun 01, 2021 at 04:11:08PM +0800, Zhu Yanjun wrote:
On Tue, Jun 1, 2021 at 3:56 PM kamal heib <kamalheib1@xxxxxxxxx> wrote:
On 1 Jun 2021, at 10:45, Zhu Yanjun <zyjzyj2000@xxxxxxxxx> wrote:
On Tue, Jun 1, 2021 at 1:58 PM Kamal Heib <kamalheib1@xxxxxxxxx> wrote:
To avoid the following failure when trying to load the rdma_rxe module
while IPv6 is disabled, Add a check to make sure that IPv6 is enabled
before trying to create the IPv6 UDP tunnel.
$ modprobe rdma_rxe
modprobe: ERROR: could not insert 'rdma_rxe': Operation not permitted
About this problem, this link:
https://patchwork.kernel.org/project/linux-rdma/patch/20210413234252.12209-1-yanjun.zhu@xxxxxxxxx/
also tries to solve ipv6 problem.
Zhu Yanjun
Yes, but this patch is fixing the problem more cleanly and I’ve tested it.
Please check this link
https://lore.kernel.org/linux-rdma/20210326012723.41769-1-yanjun.zhu@xxxxxxxxx/T/
carefully.
Please pay attention to the comments from Jason Gunthorpe
I think the comment still holds, the correct fix is to detect the -97
errno down in the call chain and just ignore ipv6 support in this
case.
Jason
OK, Could you please tell me what do you think about the following:
diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index 01662727dca0..144d9e1c1c3d 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -208,6 +208,11 @@ static struct socket *rxe_setup_udp_tunnel(struct net *net, __be16 port,
/* Create UDP socket */
err = udp_sock_create(net, &udp_cfg, &sock);
if (err < 0) {
+ if (ipv6 && (err == -EAFNOSUPPORT)) {
+ pr_warn("IPv6 is not supported can not create UDP socket\n");
+ return NULL;
The returned value is changed.
Zhu Yanjun
+ }
+
pr_err("failed to create udp socket. err = %d\n", err);
return ERR_PTR(err);
}
Thanks,
Kamal
