Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20181120105124.14733-1-christian@brauner.io> <87in0g5aqo.fsf@oldenburg.str.redhat.com>
 <36323361-90BD-41AF-AB5B-EE0D7BA02C21@amacapital.net> <993B98AC-51DF-4131-AF7F-7DA2A7F485F1@brauner.io>
 <CALCETrWnQNMQcCmFZrftVVYgAMW6DT3gyxvVb_v9_enUCUkHiw@mail.gmail.com>
 <20181129195551.woe2bl3z3yaysqb6@brauner.io> <6E21165F-2C76-4877-ABD9-0C86D55FD6AA@amacapital.net>
 <CAK8P3a2i9DFaS9-3A9V_pKwFfwVR0bDqgLosmivge3Cx2VyiuQ@mail.gmail.com>
 <87y39b2lm2.fsf@xmission.com> <20181130065606.kmilbbq46oeycjp5@brauner.io>
 <CAK8P3a0kqPii5TwFAo_JHLX=o_FDMFVKXxgzzbDjLFZ7OQ5QCQ@mail.gmail.com>
 <87y399s3sc.fsf@xmission.com> <87tvjxp8pc.fsf@xmission.com>
 <F53D6D38-3521-4C20-9034-5AF447DF62FF@amacapital.net> <87lg58pzae.fsf@xmission.com>
In-Reply-To: <87lg58pzae.fsf@xmission.com>
From:   Andy Lutomirski <luto@kernel.org>
Date:   Sat, 1 Dec 2018 17:14:45 -0800
Message-ID: <CALCETrUsHt=ft8Huy3NUn4uKqoys83aArHBAPACFG=zCGibC6A@mail.gmail.com>
Subject: Re: [PATCH v2] signal: add procfd_signal() syscall
To:     "Eric W. Biederman" <ebiederm@xmission.com>
Cc:     Arnd Bergmann <arnd@arndb.de>,
        Christian Brauner <christian@brauner.io>,
        Andrew Lutomirski <luto@kernel.org>,
        Florian Weimer <fweimer@redhat.com>,
        LKML <linux-kernel@vger.kernel.org>,
        "Serge E. Hallyn" <serge@hallyn.com>, Jann Horn <jannh@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Oleg Nesterov <oleg@redhat.com>,
        Aleksa Sarai <cyphar@cyphar.com>,
        Al Viro <viro@zeniv.linux.org.uk>,
        Linux FS Devel <linux-fsdevel@vger.kernel.org>,
        Linux API <linux-api@vger.kernel.org>,
        Daniel Colascione <dancol@google.com>,
        Tim Murray <timmurray@google.com>,
        linux-man <linux-man@vger.kernel.org>,
        Kees Cook <keescook@chromium.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sat, Dec 1, 2018 at 4:07 PM Eric W. Biederman <ebiederm@xmission.com> wr=
ote:
>
> Andy Lutomirski <luto@amacapital.net> writes:
>
> >> On Dec 1, 2018, at 7:28 AM, Eric W. Biederman <ebiederm@xmission.com> =
wrote:
> >>
> >>
> >> It just occurs to me that the simple way to implement
> >> procfd_sigqueueinfo info is like:
> >>
> >> int copy_siginfo_from_user_any(kernel_siginfo_t *info, siginfo_t *uinf=
o)
> >> {
> >> #ifdef CONFIG_COMPAT
> >>    if (in_compat_syscall)
> >>        return copy_siginfo_from_user32(info, uinfo);
> >> #endif
> >>    return copy_siginfo_from_user(info, uinfo);
> >> }
> >>
> >> long procfd_sigqueueinfo(int fd, siginfo_t *uinfo)
> >> {
> >>    kernel_siginfo info;
> >>
> >>        if (copy_siginfo_from_user_any(&info, uinfo))
> >>            return -EFAULT;
> >>    ...;
> >> }
> >>
> >> It looks like there is already a place in ptrace.c that already
> >> hand rolls copy_siginfo_from_user_any.
> >>
> >> So while I would love to figure out the subset of siginfo_t tha we can
> >> just pass through, as I think that would make a better more forward
> >> compatible copy_siginfo_from_user32.
> >
> > Seems reasonable to me. It=E2=80=99s less code overall than any other s=
uggestion, too.
> >
> >>  I think for this use case we just
> >> add the in_compat_syscall test and then we just need to ensure this ne=
w
> >> system call is placed in the proper places in the syscall table.
> >>
> >> Because we will need 3 call sights: x86_64, x32 and ia32.  As the layo=
ut
> >> changes between those three subarchitecuters.
> >>
> >>
> >
> > If it=E2=80=99s done this way, it can just be =E2=80=9Ccommon=E2=80=9D =
in the 64-bit
> > table. And we kick the can a bit farther down the road :)
> >
> > I=E2=80=99m working on patches to clean up x86=E2=80=99s syscall mess. =
It=E2=80=99s slow
> > because I keep finding new messes.  So far I have rt_sigreturn working
> > like every other syscall =E2=80=94 whee.
> >
> > Also, Eric, for your edification, I have a draft patch set to
> > radically simplify x86=E2=80=99s signal delivery and return.  Once that=
=E2=80=99s
> > done, I can trivially speed up delivery by a ton by using sysret.
>
> Nice.
>
> Do we care about the performance of synchronous signal delivery (AKA
> hardware exceptions) vs ordinary signal delivery.  I get the feeling
> there are serious simplifications to be had in that case.
>


I dunno what user code cares about.  Linux's support for synchronous
exception handling is so far behind, say, Windows, that I don't know
if it's even used for anything very serious.  We should probably
profile it after I finish my changes and we can see how bad it is.  We
can't do anything at all about the time it takes the CPU to deliver
the exception, and trying to avoid IRET when we return would be tricky
at best, although siglongjmp() might end up skipping it.