On Thu, Sep 27, 2018 at 2:30 PM Vladis Dronov [off-list ref] wrote:
Hello,
This report is actually for the same bug which was reported in:
https://syzkaller.appspot.com/bug?id=088efeac32fdde781038a777a63e436c0d4d7036
The note there that the bug was fixed by "Commits: net: fix uninit-value in
__hw_addr_add_ex()" is wrong. A C-reproducer from the 2nd syzkaller report
can trigger the bug from this one.
I've researched this and a result is a proposed patch, the problem is the tun
device code allowing to set an arbitrary link type.
https://lkml.org/lkml/2018/9/26/416
https://lore.kernel.org/lkml/20180926093018.6646-1-vdronov@redhat.com/T/#u (local)
https://marc.info/?l=linux-netdev&m=153795423320016&w=2
I dunno, your patch looks quite not the right fix.
If TUN is able to change dev->type, how comes it does not set the
appropriate dev->addr_len at the same time ?
Really the bug seems to be deeper, and without setting proper
dev->addr_len, we'll need more 'fixes' like yours.
Thanks.