Re: [PATCH rdma-next v3 04/14] RDMA/nldev: Report net namespace move errors through extack

[Jiri Pirko <jiri@resnulli.us>]

Fri, Jul 17, 2026 at 03:24:25PM +0200, sashiko-bot@kernel.org wrote:
>Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
>- [High] The newly added feature to support renaming an RDMA device while moving its network namespace is inaccessible from userspace due to a logic flaw in the netlink handler. Additionally, providing both attributes silently aborts the namespace move.
>--
>
>--- Patch [4]: [PATCH rdma-next v3 04/14] RDMA/nldev: Report net namespace move errors through extack ---
>commit dd68e4a951ab0b58504bc034673213514633a1a1
>Author: Jiri Pirko <jiri@resnulli.us>
>
>RDMA/nldev: Report net namespace move errors through extack
>
>Thread extack through the existing net namespace move helper and report the
>main failure reasons from the core path. Keep the existing move UAPI shape
>unchanged.
>
>> diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
>> index 77a7580801486..8648e95700bfa 100644
>> --- a/drivers/infiniband/core/nldev.c
>> +++ b/drivers/infiniband/core/nldev.c
>> @@ -1195,10 +1195,8 @@ static int nldev_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
>>  		u32 ns_fd;
>>  
>>  		ns_fd = nla_get_u32(tb[RDMA_NLDEV_NET_NS_FD]);
>> -		err = ib_device_set_netns_put(skb, device, ns_fd, NULL);
>> -		if (err == -EEXIST)
>> -			NL_SET_ERR_MSG(extack,
>> -				       "Device name already exists in the target net namespace");
>> +		err = ib_device_set_netns_put(skb, device, ns_fd, NULL,
>> +					      extack);
>
>[Severity: High]
>Does this logic prioritize the device rename and exit early, preventing the
>namespace move from occurring if both attributes are present?
>
>Looking at nldev_set_doit(), if a user with CAP_NET_ADMIN provides both
>RDMA_NLDEV_ATTR_DEV_NAME and RDMA_NLDEV_NET_NS_FD in a single message:
>
>    if (tb[RDMA_NLDEV_ATTR_DEV_NAME]) {
>        /* ... */
>        err = ib_device_rename(device, name);
>        goto done;
>    }
>
>won't the handler process the rename and hit a goto done; silently skipping
>the namespace move? This could result in a success response while leaving the
>device in the old namespace.
>
>Additionally, it unconditionally passes NULL as the name to
>ib_device_set_netns_put() in the patched hunk above. Doesn't this render
>the rename-on-move functionality completely inaccessible from userspace?

Both concerns are resolved by the follow-up patch.