Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
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: [RFC] Get siginfo from unreaped task
Date:   Sat, 12 Feb 2022 18:32:08 -0800
Message-Id: <CF5167CE-FA1C-4CEC-9EA8-5EE8041FE7C4@amacapital.net>
References: <CAP145phC6S6Zda-ZWLH1s4ZfDPh79rtf_7vzs-yvt1vykUCP4A@mail.gmail.com>
Cc:     Kees Cook <keescook@chromium.org>,
        "Eric W. Biederman" <ebiederm@xmission.com>,
        Jann Horn <jannh@google.com>, Will Drewry <wad@chromium.org>,
        linux-api@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-hardening@vger.kernel.org
In-Reply-To: <CAP145phC6S6Zda-ZWLH1s4ZfDPh79rtf_7vzs-yvt1vykUCP4A@mail.gmail.com>
To:     =?utf-8?Q?Robert_=C5=9Awi=C4=99cki?= <robert@swiecki.net>
Precedence: bulk


> On Feb 12, 2022, at 3:24 AM, Robert =C5=9Awi=C4=99cki <robert@swiecki.net>=
 wrote:
>=20
> =EF=BB=BFsob., 12 lut 2022 o 05:28 Kees Cook <keescook@chromium.org> napis=
a=C5=82(a):
>>=20
>> Make siginfo available through PTRACE_GETSIGINFO after process death,
>> without needing to have already used PTRACE_ATTACH. Uses 48 more bytes
>> in task_struct, though I bet there might be somewhere else we could
>> stash a copy of it?
>=20
> An alternative way of accessing this info could be abusing the
> waitid() interface, with some additional, custom to Linux, flag
>=20
> waitid(P_ALL, 0, &si, __WCHILDSIGINFO);
>=20
> which would change what is put into si.
>=20
> But maybe ptrace() is better, because it's mostly incompatible with
> other OSes anyway on the behavior/flag level, while waitd() seems to
> be POSIX/BSD standard, even if Linux specifies some additional flags.
>=20
>=20

I had a kind of opposite thought, which is that it would be very nice to be a=
ble to get all the waitid() data without reaping a process or even necessari=
ly being its parent.  Maybe these can be combined?  A new waitid() option li=
ke you=E2=80=99re suggesting could add siginfo (and might need permissions).=
  And we could have a different waitid() flag that says =E2=80=9Cmaybe not m=
y child, don=E2=80=99t reap=E2=80=9D (and also needs permissions).

Although the =E2=80=9Cdon=E2=80=99t reap=E2=80=9D thing is fundamentally rac=
y. What a sane process manager actually wants is an interface to read all th=
is info from a pidfd, which means it all needs to get stuck in struct pid. A=
nd task_struct needs a completion or wait queue so you can actually wait for=
 a pidfd to exit (unless someone already did this =E2=80=94 I had patches a w=
hile back).  And this would be awesome.=