Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20201026125227.54520-1-a.nogikh@gmail.com> <20201026125227.54520-3-a.nogikh@gmail.com>
 <CAC5umyjZn+gtO=u8GhptVggzodbY-vnk9NNTOqyZyUW2up2ASQ@mail.gmail.com> <CANp29Y7sDaxrw+0wEVU0vo56AhKFRvbQSGeRc168gs6S8iK_-g@mail.gmail.com>
In-Reply-To: <CANp29Y7sDaxrw+0wEVU0vo56AhKFRvbQSGeRc168gs6S8iK_-g@mail.gmail.com>
From:   Dmitry Vyukov <dvyukov@google.com>
Date:   Wed, 28 Oct 2020 10:41:04 +0100
Message-ID: <CACT4Y+ZX=FjO0Ohoxnyjb3RqaTdGDpYs-Z4pJyiTo2TYY_ROqQ@mail.gmail.com>
Subject: Re: [RFC PATCH v2 2/2] docs: add fail_lsm_hooks info to fault-injection.rst
To:     Aleksandr Nogikh <nogikh@google.com>
Cc:     Akinobu Mita <akinobu.mita@gmail.com>,
        Aleksandr Nogikh <a.nogikh@gmail.com>,
        James Morris <jmorris@namei.org>,
        "Serge E. Hallyn" <serge@hallyn.com>,
        Andrey Konovalov <andreyknvl@google.com>,
        Marco Elver <elver@google.com>,
        Alexander Potapenko <glider@google.com>,
        Kees Cook <keescook@google.com>,
        LKML <linux-kernel@vger.kernel.org>,
        linux-security-module <linux-security-module@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Oct 27, 2020 at 6:34 PM Aleksandr Nogikh <nogikh@google.com> wrote:
> [...]
> > In addition to this global one, what do you think about per-hook fault
> > injection,
> > i.e. /sys/kernel/debug/fail_lsm_hooks/<FUNC>/retval ?
>
> I was thinking about this, but decided to begin with a simple version
> that could definitely be useful in practice (for syzbot/syzkaller it is just
> necessary to have a fault injection capability that will be triggered via
> fail-nth). If per-hook fault injection can also be useful to someone, I
> can try to add it as well.

Yes, before we add it, it would be useful to have a clear use case
(otherwise we can add an unused thing, or implement it in a way that
slightly misses the use case).
Note that fail-nth allows to fail a single concrete site for testing,
though it's not super convenient for this as one would need to figure
out the right N first. But as a one-off test it should do.


> > In this case, we need a fault_attr for each hook. (Maybe, we can use the same
> > technique that is used to define security_hook_heads).
>
> Yes, that technique should help to implement the feature in a very concise
> way. Thanks for the suggestion.