cfg80211: always free wiphy specific regdomain
authorJohannes Berg <johannes.berg@intel.com>
Mon, 27 Sep 2021 11:11:06 +0000 (13:11 +0200)
committerJohannes Berg <johannes.berg@intel.com>
Mon, 27 Sep 2021 11:30:33 +0000 (13:30 +0200)
In the (somewhat unlikely) event that we allocate a wiphy, then
add a regdomain to it, and then fail registration, we leak the
regdomain. Fix this by just always freeing it at the end, in the
normal cases we'll free (and NULL) it during wiphy_unregister().
This happened when the wiphy settings were bad, and since they
can be controlled by userspace with hwsim, syzbot was able to
find this issue.

Reported-by: syzbot+1638e7c770eef6b6c0d0@syzkaller.appspotmail.com
Fixes: 3e0c3ff36c4c ("cfg80211: allow multiple driver regulatory_hints()")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20210927131105.68b70cef4674.I4b9f0aa08c2af28555963b9fe3d34395bb72e0cc@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
net/wireless/core.c

index 03323121ca505c0d7c46a1cbfde94d1c2cfc9c5c..45be124a98f156ca46c99ef8ad288e8a411995fc 100644 (file)
@@ -1080,6 +1080,16 @@ void cfg80211_dev_free(struct cfg80211_registered_device *rdev)
        list_for_each_entry_safe(scan, tmp, &rdev->bss_list, list)
                cfg80211_put_bss(&rdev->wiphy, &scan->pub);
        mutex_destroy(&rdev->wiphy.mtx);
+
+       /*
+        * The 'regd' can only be non-NULL if we never finished
+        * initializing the wiphy and thus never went through the
+        * unregister path - e.g. in failure scenarios. Thus, it
+        * cannot have been visible to anyone if non-NULL, so we
+        * can just free it here.
+        */
+       kfree(rcu_dereference_raw(rdev->wiphy.regd));
+
        kfree(rdev);
 }