On Thu, Sep 3, 2020 at 2:47 AM Eric Dumazet <eric.duma...@gmail.com> wrote: > This commit might be related : > > commit 4e407ff5cd67ec76eeeea1deec227b7982dc7f66 > Author: Cong Wang <xiyou.wangc...@gmail.com> > Date: Sun Aug 19 12:22:12 2018 -0700 > > act_ife: move tcfa_lock down to where necessary
It does not look like my commit's fault. From my _quick_ understanding of this problem, we somehow have a "deadlock" situation: Thread A allocates an IDR with a specific index and calls populate_metalist() to re-accquire the RTNL lock. Thread B gets RTNL lock right after thread A releases it, then it waits for thread A to finally commit the IDR change. Thanks.
