2023-10-04 14:55:16

by Daniel Jordan

[permalink] [raw]
Subject: Re: Issue: padata: UAF in padata_serial_worker reproduce

Hello,

On Thu, Sep 21, 2023 at 08:50:17AM +0800, Wang Jinchao wrote:
> On Mon, Sep 18, 2023 at 09:21:59AM -0400, Daniel Jordan wrote:
> > Hi,
> >
> > On Mon, Sep 18, 2023 at 02:27:34PM +0800, Wang Jinchao wrote:
> > > Hello, I have reproduced the issue mentioned by Nicolai Stange on an arm64 server.
> > > After applying Nicolai Stange's patch, the issue did not reoccur even after running for two days.
> > > The test tools used were ltp20220121 and stress-ng-0.15.06.tar.gz.
> > >
> > > refurl: https://lore.kernel.org/all/[email protected]/T/#u
> >
> > Yeah, this series seems close to being finished, just hasn't made it
> > over the finish line quite yet. I wonder if Nicolai has plans to
> > ressurect or whether someone else should take it over.
>
> Hello, I'd like to inquire about something:
>
> Since users of padata_free_shell need to ensure that all tasks in padata_shell have completed,
> how can they confirm this when they are uncertain? Would it be better if Padata provided a mechanism,
> such as padata_shell_status(), to maintain statistics like the number of tasks?

Looks like you found a way here:

https://lore.kernel.org/all/ZRTLHY5A+VqIKhA2@fedora/