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: pt_regs->ax == -ENOSYS
Date:   Tue, 27 Apr 2021 16:51:06 -0700
Message-Id: <78327AF2-575F-415E-98D4-6ECDFE311D0F@amacapital.net>
References: <202104271619.0DBE456@keescook>
Cc:     "H. Peter Anvin" <hpa@zytor.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Ingo Molnar <mingo@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Andrew Lutomirski <luto@kernel.org>,
        Borislav Petkov <bp@alien8.de>, linux-kernel@vger.kernel.org,
        oleg@redhat.com, Will Drewry <wad@chromium.org>
In-Reply-To: <202104271619.0DBE456@keescook>
To:     Kees Cook <keescook@chromium.org>
Precedence: bulk


> On Apr 27, 2021, at 4:29 PM, Kees Cook <keescook@chromium.org> wrote:
>=20
> =EF=BB=BFOn Tue, Apr 27, 2021 at 03:58:03PM -0700, H. Peter Anvin wrote:
>>> On 4/27/21 2:28 PM, Andy Lutomirski wrote:
>>>=20
>>>> On Apr 27, 2021, at 2:15 PM, H. Peter Anvin <hpa@zytor.com> wrote:
>>>>=20
>>>> =EF=BB=BFTrying to stomp out some possible cargo cult programming?
>>>>=20
>>>> In the process of going through the various entry code paths, I have to=
 admit to being a bit confused why pt_regs->ax is set to -ENOSYS very early i=
n the system call path.
>>>>=20
>>>=20
>>> It has to get set to _something_, and copying orig_ax seems perhaps sill=
y.  There could also be code that relies on ptrace poking -1 into the nr res=
ulting in -ENOSYS.
>>>=20
>>=20
>> Yeah. I obviously ran into this working on the common entry-exit code for=

>> FRED; the frame has annoyingly different formats because of this, and I
>> wanted to avoid slowing down the system call path.
>>=20
>>>> What is perhaps even more confusing is:
>>>>=20
>>>> __visible noinstr void do_syscall_64(struct pt_regs *regs, unsigned lon=
g nr)
>>>> {
>>>>        nr =3D syscall_enter_from_user_mode(regs, nr);
>>>>=20
>>>>        instrumentation_begin();
>>>>        if (likely(nr < NR_syscalls)) {
>>>>                nr =3D array_index_nospec(nr, NR_syscalls);
>>>>                regs->ax =3D sys_call_table[nr](regs);
>>>> #ifdef CONFIG_X86_X32_ABI
>>>>        } else if (likely((nr & __X32_SYSCALL_BIT) &&
>>>>                          (nr & ~__X32_SYSCALL_BIT) < X32_NR_syscalls)) {=

>>>>                nr =3D array_index_nospec(nr & ~__X32_SYSCALL_BIT,
>>>>                                        X32_NR_syscalls);
>>>>                regs->ax =3D x32_sys_call_table[nr](regs);
>>>> #endif
>>>>        }
>>>>        instrumentation_end();
>>>>        syscall_exit_to_user_mode(regs);
>>>> }
>>>> #endif
>>>>=20
>>>> Now, unless I'm completely out to sea, it seems to me that if syscall_e=
nter_from_user_mode() changes the system call number to an invalid number an=
d pt_regs->ax to !-ENOSYS then the system call will return a different value=
(!) depending on if it is out of range for the table (whatever was poked int=
o pt_regs->ax) or if it corresponds to a hole in the table. This seems to me=
 at least to be The Wrong Thing.
>>>=20
>>> I think you=E2=80=99re right.
>>>=20
>>>>=20
>>>> Calling regs->ax =3D sys_ni_syscall() in an else clause would arguably b=
e the right thing here, except possibly in the case where nr (or (int)nr, se=
e below) =3D=3D -1 or < 0.
>>>=20
>>> I think the check should be -1 for 64 bit but (u32)nr =3D=3D (u32)-1 for=
 the 32-bit path. Does that seem reasonable?
>=20
> FWIW, there is some confusion with how syscall_trac_enter() signals the
> "skip syscall" condition (-1L), vs actually calling "syscall -1".

Fortunately there is not, and never will be, a syscall -1.  But I agree that=
 calling max syscall + 1 should behave identically to calling a nonexistent s=
yscall in the middle of the table.=