netdev - Re: [bug report] devlink: Move devlink dev reload code to dev

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <aR8LKLebnQFnEDU6@stanley.mountain>
Date: Thu, 20 Nov 2025 15:35:52 +0300
From: Dan Carpenter <dan.carpenter@...aro.org>
To: Jiri Pirko <jiri@...nulli.us>
Cc: Moshe Shemesh <moshe@...dia.com>, netdev@...r.kernel.org
Subject: Re: [bug report] devlink: Move devlink dev reload code to dev

On Thu, Nov 20, 2025 at 01:10:39PM +0100, Jiri Pirko wrote:
> Wed, Nov 19, 2025 at 07:15:52PM +0100, dan.carpenter@...aro.org wrote:
> >On Wed, Nov 19, 2025 at 06:19:18PM +0100, Jiri Pirko wrote:
> >> Wed, Nov 19, 2025 at 09:55:58AM +0100, dan.carpenter@...aro.org wrote:
> >> >Hello Moshe Shemesh,
> >> >
> >> >Commit c6ed7d6ef929 ("devlink: Move devlink dev reload code to dev")
> >> >from Feb 2, 2023 (linux-next), leads to the following Smatch static
> >> >checker warning:
> >> >
> >> >	net/devlink/dev.c:408 devlink_netns_get()
> >> >	error: potential NULL/IS_ERR bug 'net'
> >> >
> >> >net/devlink/dev.c
> >> >    378 static struct net *devlink_netns_get(struct sk_buff *skb,
> >> >    379                                      struct genl_info *info)
> >> >    380 {
> >> >    381         struct nlattr *netns_pid_attr = info->attrs[DEVLINK_ATTR_NETNS_PID];
> >> >    382         struct nlattr *netns_fd_attr = info->attrs[DEVLINK_ATTR_NETNS_FD];
> >> >    383         struct nlattr *netns_id_attr = info->attrs[DEVLINK_ATTR_NETNS_ID];
> >> >    384         struct net *net;
> >> >    385 
> >> >    386         if (!!netns_pid_attr + !!netns_fd_attr + !!netns_id_attr > 1) {
> >> >    387                 NL_SET_ERR_MSG(info->extack, "multiple netns identifying attributes specified");
> >> >    388                 return ERR_PTR(-EINVAL);
> >> >    389         }
> >> >    390 
> >> >    391         if (netns_pid_attr) {
> >> >    392                 net = get_net_ns_by_pid(nla_get_u32(netns_pid_attr));
> >> >
> >> >Smatch thinks that the "net = get_net(nsproxy->net_ns);" could mean that
> >> >get_net_ns_by_pid() returns NULL.  I don't know if that's correct or not.
> >> >If someone could tell me, then it's easy for me to add a line
> >> >"get_net_ns_by_pid 0" to the smatch_data/db/kernel.delete.return_states
> >> >file but I'd prefer to be sure before I do that...
> >> 
> >> I don't see how get_net() can return NULL.
> >> 
> >
> >It returns whatever you pass to it.  The ns_ref_inc() macro
> >has a NULL check built in so it accepts NULL pointers.
> 
> Of course, I don't see how NULL can be passed to in in the current code.
> 

Fair enough...  This is a complicated thing to track with static analysis
because create_new_namespaces() has a bit of recursion as far a Smatch is
concerned.

	new_nsp->net_ns = copy_net_ns(flags, user_ns, tsk->nsproxy->net_ns);

It's copying the old tsk->nsproxy->net_ns to new_nsp->net_ns and Smatch
says if the old one is NULL then the new one could be NULL too.  It's
just a bit complicated is all.

Anyway, it's a oneliner fix in Smatch.

regards,
dan carpenter