Re: [PATCH -rcu] asm-generic, kcsan: Add KCSAN instrumentation for bitops

[Marco Elver <elver@google.com>]

On Mon, 20 Jan 2020 at 20:03, Arnd Bergmann <arnd@arndb.de> wrote:
>
> On Mon, Jan 20, 2020 at 4:11 PM Marco Elver <elver@google.com> wrote:
> > On Mon, 20 Jan 2020 at 15:40, Arnd Bergmann <arnd@arndb.de> wrote:
> > > On Mon, Jan 20, 2020 at 3:23 PM Marco Elver <elver@google.com> wrote:
> > > > On Fri, 17 Jan 2020 at 14:14, Marco Elver <elver@google.com> wrote:
> > > > > On Fri, 17 Jan 2020 at 13:25, Arnd Bergmann <arnd@arndb.de> wrote:
> > > > > > On Wed, Jan 15, 2020 at 9:50 PM Marco Elver <elver@google.com> wrote:
> > >
> > > > > > If you can't find any, I would prefer having the simpler interface
> > > > > > with just one set of annotations.
> > > > >
> > > > > That's fair enough. I'll prepare a v2 series that first introduces the
> > > > > new header, and then applies it to the locations that seem obvious
> > > > > candidates for having both checks.
> > > >
> > > > I've sent a new patch series which introduces instrumented.h:
> > > >    http://lkml.kernel.org/r/20200120141927.114373-1-elver@google.com
> > >
> > > Looks good to me, feel free to add
> > >
> > > Acked-by: Arnd Bergmann <arnd@arndb.de>
> > >
> > > if you are merging this through your own tree or someone else's,
> > > or let me know if I should put it into the asm-generic git tree.
> >
> > Thank you!  It seems there is still some debate around the user-copy
> > instrumentation.
> >
> > The main question we have right now is if we should add pre/post hooks
> > for them. Although in the version above I added KCSAN checks after the
> > user-copies, it seems maybe we want it before. I personally don't have
> > a strong preference, and wanted to err on the side of being more
> > conservative.
> >
> > If I send a v2, and it now turns out we do all the instrumentation
> > before the user-copies for KASAN and KCSAN, then we have a bunch of
> > empty hooks. However, for KMSAN we need the post-hook, at least for
> > copy_from_user. Do you mind a bunch of empty functions to provide
> > pre/post hooks for user-copies? Could the post-hooks be generally
> > useful for something else?
>
> I'd prefer not to add any empty hooks, let's do that once they
> are actually used.

I hope I found a solution to the various constraints:
http://lkml.kernel.org/r/20200121160512.70887-1-elver@google.com

I removed your Acks from the patches that were changed in v2. Please
have another look.

Re tree: Once people are happy with the patches, since this depends on
KCSAN it'll probably have to go through Paul's -rcu tree, since KCSAN
is not yet in mainline (currently only in -rcu, -tip, and -next).

Thanks,
-- Marco