Re: [GIT PULL] SafeSetID LSM changes for 5.4

From: Micah Morton <mortonm@chromium.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: James Morris <jamorris@linuxonhyperv.com>,
	Jann Horn <jannh@google.com>,
	Bart Van Assche <bart.vanassche@wdc.com>,
	"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	linux-security-module <linux-security-module@vger.kernel.org>
Subject: Re: [GIT PULL] SafeSetID LSM changes for 5.4
Date: Mon, 23 Sep 2019 20:30:54 -0700	[thread overview]
Message-ID: <CAJ-EccMy=tNPp3=PQZxLT7eovojoAdpfQmqhAyv7XO3GwPQBMg@mail.gmail.com> (raw)
In-Reply-To: <CAHk-=wh4cuHsE8jFHO7XVatdXa=M2f4RHL3VwnSkAf5UNHUJ-Q@mail.gmail.com>

On Mon, Sep 23, 2019 at 5:45 PM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> On Mon, Sep 23, 2019 at 4:35 PM James Morris <jamorris@linuxonhyperv.com> wrote:
> >
> > My understanding is that SafeSetID is shipping in ChromeOS -- this was
> > part of the rationale for merging it.
>
> Well, if even the developer didn't test it for two months, I don't
> think "it's in upstream" makes any sense or difference.
>
>                      Linus

Yes, SafeSetID is shipping on Chrome OS, although I agree having that
bug in 5.3 without anyone noticing is bad. When Jann sent the last
round of patches for 5.3 he had tested the code and everything looked
good, although I unfortunately neglected to test it again after a
tweak to one of the patches, which of course broke stuff when the
patches ultimately went in.

Even though this is enabled in production for Chrome OS, none of the
Chrome OS devices are using version 5.3 yet, so it went unnoticed on
Chrome OS so far. In general the fact that a kernel feature is
shipping on Chrome OS isn't an up-to-date assurance that the feature
works in the most recent Linux release, as it would likely be months
(at least) from when a change makes it into the kernel until that
kernel release is ever run on a Chrome OS device (right now the most
recent kernel we ship on Chrome OS is 4.19, so I've had to backport
the SafeSetID stuff).

We've found this SafeSetID LSM to be pretty useful on Chrome OS, and
more use cases have popped up than we had in mind when writing it,
which suggests others would potentially find it useful as well. But I
understand for it to be useful to others it needs to be stable and
functional on every release. The best way I know of ensuring this is
for me to personally run the SafeSetID selftest (in
tools/testing/selftests/safesetid/) every release, regardless of
whether we make any changes to SafeSetID itself. Does this sound
sufficient or are there more formal guidelines/processes here that I'm
not aware of?