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;
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
From:   Andy Lutomirski <luto@amacapital.net>
Mime-Version: 1.0 (1.0)
Subject: Re: [PATCH v2 seccomp 3/6] seccomp/cache: Add "emulator" to check if filter is arg-dependent
Date:   Fri, 25 Sep 2020 17:34:29 -0700
Message-Id: <677FA6F9-D577-4594-9FDC-D70B0D6900C6@amacapital.net>
References: <202009251648.4AA27D5B@keescook>
Cc:     YiFei Zhu <zhuyifei1999@gmail.com>,
        Linux Containers <containers@lists.linux-foundation.org>,
        YiFei Zhu <yifeifz2@illinois.edu>, bpf <bpf@vger.kernel.org>,
        kernel list <linux-kernel@vger.kernel.org>,
        Aleksa Sarai <cyphar@cyphar.com>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Dimitrios Skarlatos <dskarlat@cs.cmu.edu>,
        Giuseppe Scrivano <gscrivan@redhat.com>,
        Hubertus Franke <frankeh@us.ibm.com>,
        Jack Chen <jianyan2@illinois.edu>,
        Jann Horn <jannh@google.com>,
        Josep Torrellas <torrella@illinois.edu>,
        Tianyin Xu <tyxu@illinois.edu>,
        Tobin Feldman-Fitzthum <tobin@ibm.com>,
        Tycho Andersen <tycho@tycho.pizza>,
        Valentin Rothberg <vrothber@redhat.com>,
        Will Drewry <wad@chromium.org>
In-Reply-To: <202009251648.4AA27D5B@keescook>
To:     Kees Cook <keescook@chromium.org>
Precedence: bulk


> On Sep 25, 2020, at 4:49 PM, Kees Cook <keescook@chromium.org> wrote:
>=20
> =EF=BB=BFOn Fri, Sep 25, 2020 at 02:07:46PM -0700, Andy Lutomirski wrote:
>>> On Fri, Sep 25, 2020 at 1:37 PM Kees Cook <keescook@chromium.org> wrote:=

>>>=20
>>> On Fri, Sep 25, 2020 at 12:51:20PM -0700, Andy Lutomirski wrote:
>>>>=20
>>>>=20
>>>>> On Sep 25, 2020, at 12:42 PM, Kees Cook <keescook@chromium.org> wrote:=

>>>>>=20
>>>>> =EF=BB=BFOn Fri, Sep 25, 2020 at 11:45:05AM -0500, YiFei Zhu wrote:
>>>>>> On Thu, Sep 24, 2020 at 10:04 PM YiFei Zhu <zhuyifei1999@gmail.com> w=
rote:
>>>>>>>> Why do the prepare here instead of during attach? (And note that it=

>>>>>>>> should not be written to fail.)
>>>>>>>=20
>>>>>>> Right.
>>>>>>=20
>>>>>> During attach a spinlock (current->sighand->siglock) is held. Do we
>>>>>> really want to put the emulator in the "atomic section"?
>>>>>=20
>>>>> It's a good point, but I had some other ideas around it that lead to m=
e
>>>>> a different conclusion. Here's what I've got in my head:
>>>>>=20
>>>>> I don't view filter attach (nor the siglock) as fastpath: the lock is
>>>>> rarely contested and the "long time" will only be during filter attach=
.
>>>>>=20
>>>>> When performing filter emulation, all the syscalls that are already
>>>>> marked as "must run filter" on the previous filter can be skipped for
>>>>> the new filter, since it cannot change the outcome, which makes the
>>>>> emulation step faster.
>>>>>=20
>>>>> The previous filter's bitmap isn't "stable" until siglock is held.
>>>>>=20
>>>>> If we do the emulation step before siglock, we have to always do full
>>>>> evaluation of all syscalls, and then merge the bitmap during attach.
>>>>> That means all filters ever attached will take maximal time to perform=

>>>>> emulation.
>>>>>=20
>>>>> I prefer the idea of the emulation step taking advantage of the bitmap=

>>>>> optimization, since the kernel spends less time doing work over the li=
fe
>>>>> of the process tree. It's certainly marginal, but it also lets all the=

>>>>> bitmap manipulation stay in one place (as opposed to being split betwe=
en
>>>>> "prepare" and "attach").
>>>>>=20
>>>>> What do you think?
>>>>>=20
>>>>>=20
>>>>=20
>>>> I=E2=80=99m wondering if we should be much much lazier. We could potent=
ially wait until someone actually tries to do a given syscall before we try t=
o evaluate whether the result is fixed.
>>>=20
>>> That seems like we'd need to track yet another bitmap of "did we emulate=

>>> this yet?" And it means the filter isn't really "done" until you run
>>> another syscall? eeh, I'm not a fan: it scratches at my desire for
>>> determinism. ;) Or maybe my implementation imagination is missing
>>> something?
>>>=20
>>=20
>> We'd need at least three states per syscall: unknown, always-allow,
>> and need-to-run-filter.
>>=20
>> The downsides are less determinism and a bit of an uglier
>> implementation.  The upside is that we don't need to loop over all
>> syscalls at load -- instead the time that each operation takes is
>> independent of the total number of syscalls on the system.  And we can
>> entirely avoid, say, evaluating the x32 case until the task tries an
>> x32 syscall.
>>=20
>> I think it's at least worth considering.
>=20
> Yeah, worth considering. I do still think the time spent in emulation is
> SO small that it doesn't matter running all of the syscalls at attach
> time. The filters are tiny and fail quickly if anything "interesting"
> start to happen. ;)
>=20

There=E2=80=99s a middle ground, too: do it lazily per arch.  So we would al=
locate and populate the compat bitmap the first time a compat syscall is att=
empted and do the same for x32. This may help avoid the annoying extra memor=
y usage and 3x startup overhead while retaining full functionality.=