Re: [PATCH 4.4 40/57] tpm: Provide strong locking for device removal

From: Jarkko Sakkinen
Date: Mon Jul 31 2017 - 18:23:04 EST


On Fri, Jul 28, 2017 at 03:42:18PM -0700, Greg Kroah-Hartman wrote:
> On Wed, Jul 26, 2017 at 02:03:06PM -0600, Jason Gunthorpe wrote:
> > On Wed, Jul 26, 2017 at 12:56:37PM -0700, Greg Kroah-Hartman wrote:
> > > On Tue, Jul 25, 2017 at 11:56:01PM +0100, Ben Hutchings wrote:
> > > > On Wed, 2017-07-19 at 13:12 +0200, Greg Kroah-Hartman wrote:
> > > > > 4.4-stable review patch. If anyone has any objections, please let me know.
> > > > >
> > > > >
> > > > > From: Jason Gunthorpe <jgunthorpe@xxxxxxxxxxxxxxxxxxxx>
> > > > >
> > > > > commit 4e26195f240d73150e8308ae42874702e3df8d2c upstream.
> > > > >
> > > > > Add a read/write semaphore around the ops function pointers so
> > > > > ops can be set to null when the driver un-registers.
> > > > [...]
> > > > > @@ -49,10 +99,10 @@ struct tpm_chip *tpm_chip_find_get(int c
> > > > > if (chip_num != TPM_ANY_NUM && chip_num != pos->dev_num)
> > > > > continue;
> > > > >
> > > > > - if (try_module_get(pos->dev.parent->driver->owner)) {
> > > > > + /* rcu prevents chip from being free'd */
> > > > > + if (!tpm_try_get_ops(pos))
> > > > [...]
> > > >
> > > > But an RCU read-side critical section is an atomic context, and
> > > > semaphore operations can block! Fixed upstream by:
> > > >
> > > > commit 15516788e581eb32ec1c50e5f00aba3faf95d817
> > > > Author: Stefan Berger <stefanb@xxxxxxxxxxxxxxxxxx>
> > > > Date: Mon Feb 29 08:53:02 2016 -0500
> > > >
> > > > tpm: Replace device number bitmap with IDR
> > >
> > > Ugh, that's a big patch.
> > >
> > > Jason, Stefan, and Jarkko, what do you think? Should I also take this
> > > for 4.4-stable?
> >
> > 15516 is part of the series that included 4e26, I wouldn't take that
> > series piecemeal, as Ben observes..
> >
> > I think it would be safer to avoid all these backport patches and
> > instead restructure the important TPM shutdown patch so that it is
> > 'less safe'. This would mean there is a chance that the another TPM
> > user could send a command after shutdown, but realistically that is
> > not likely to happen.
>
> Ok, so what do you want me to do here?
>
> thanks,
>
> greg k-h

Sorry for late response. I just came from four week leave (have been
watching kernel mails only 1-2 times a week and missed this thread).

I would actually think that taking this patch would make sense as the
changes are trivial and also because this code has reminded almost
unchanged after it was added.

/Jarkko