Re: fs/pnode.c: propagate_one Oops in ltp/fs_bind test

[Christian Brauner <brauner@kernel.org>]

On Tue, Nov 15, 2022 at 11:04:01PM +0800, ditang chen wrote:
> Here is a reproducer:
> 1. Run netns.sh script in loop
> # while true; do ./netns.sh; done
> # cat netns.sh
> #!/bin/bash
> num=1000
> function create_netns()
> {
> for((i=0; i<$num; i++))
> do
>   ip netns add local$i
>   ip netns exec local$i pwd &
> done
> }
> function clean_netns()
> {
> for((i=0; i<$num; i++))
> do
>     ip netns del local$i
> done
> }
> create_netns
> clean_netns
> 
> 2.  run fs_bind/fs_bind24 in loop, fs_bind24 only
> # cat /opt/ltp/runtest/fs_bind
> #DESCRIPTION:Bind mounts and shared subtrees
> fs_bind24_sh fs_bind24.sh
> # while true; do /opt/ltp/runltp -f fs_bind; done
> 
> This oops also exists in the latest kernel code：

I've been running this since yesterday on v6.1-rc7 to reproduce and it
didn't trigger. It's unclear whether you're saying that you've managed
to reproduce this on mainline. It doesn't seem to be.